Google Inc v Trkulja [2016] VSCA 333

TORTS - Defamation - Defamatory meaning - Matter pleaded as composite publication - Whether matter relied on capable of conveying defamatory meaning.

PRACTICE AND PROCEDURE - Service of writ outside jurisdiction - Reliance on Supreme Court (General Civil Procedure) Rules 2015 r 7.01(1)(i) and (j) - Application to set aside writ and its service pursuant to Supreme Court (General Civil Procedure) Rules 2015 r 8.09(a) - Whether proceeding had no real prospect of success - Agar v Hyde (2000) 201 CLR 552 - Civil Procedure Act 2010 ss 62 and 63 - Application for leave to appeal granted and appeal allowed - Amended writ, amended statement of claim and service on defendant outside jurisdiction set aside.

1 The Court has before it an application for leave to appeal which raises difficult questions about the law of defamation as it applies to the results produced by internet search engines. At issue are the questions whether the search engine provider 'publishes' those results for the purposes of defamation law, and whether, in the particular instance, the results produced in the present matter were capable of being defamatory of the plaintiff.

2 By its further amended summons dated 12 February 2015, Google Inc ('Google' or 'the defendant') sought to set aside a defamation proceeding brought against it by Milorad[1] Trkulja ('Trkulja' or 'the plaintiff'). Pursuing its application before a judge in the Trial Division on 13 August 2015, Google relied upon the Court's power, conferred by r 8.09(a) of Chapter 1 of the Supreme Court (General Civil Procedure) Rules 2015 ('the Rules'), to set aside a writ or its service. Google had been served with the writ in the United States, the plaintiff apparently relying upon r 7.01(1)(i) and (j) of Chapter 1 of the Rules. Google had entered neither an appearance nor a conditional appearance.

3 Both below and in this Court, Google accepted, referring to Agar v Hyde,[2] that to succeed in its application it must establish that the plaintiff's proceeding has no real prospect of success. That position was correctly taken. In Agar, defendants who had been served with process overseas sought to set aside service, mainly on the footing that the plaintiffs' claims had insufficient prospects of success. Justices Gaudron, McHugh, Gummow and Hayne concluded that the test of insufficient prospects of success was no less demanding than that which operates when application is made for summary judgment. They said that -

The test to be applied has been expressed in various ways, but all of the verbal formulae which have been used are intended to describe a high degree of certainty about the ultimate outcome of the proceeding if it were allowed to go to trial in the ordinary way.[3]

the same test should be applied in deciding whether originating process served outside Australia makes claims which have such poor prospects of success that the proceeding should not go to trial as is applied in an application for summary judgment by a defendant served locally.[4]

5 In Victoria, by operation of ss 62 and 63 of the Civil Procedure Act 2010, the test is whether a plaintiff's claim 'has no real prospect of success'.

6 In support of its application, Google advanced three submissions. First, that as a matter of law it could not be held to have published the alleged defamatory matter. Second, that it would not be open to the trier of fact to conclude that the matter relied upon was defamatory of the plaintiff. Third, that, in any event, a 'search engine provider' should be immune from a proceeding such as this.

7 The judge, in reasons published on 17 November 2015, rejected all three submissions, and ordered that the summons be dismissed.

8 By application dated 9 December 2015, Google sought leave to appeal against the judge's orders; and, if leave be granted, that the appeal be allowed. On 7 October 2016, this Court heard argument respecting both the application for leave to appeal and the appeal.[5]

9 In our opinion, for the reasons which follow, Google should have leave to appeal, and the appeal should be allowed. Leave to appeal should be granted because, in our opinion, the appeal has a real, not fanciful, prospect of success.[6] The substantive appeal should be allowed because, in our opinion, the demanding test which we identified above, the burden lying upon Google, has been satisfied.

10 The only formal documents before the Court are the writ and an amended statement of claim.

11 The writ was filed on 2 December 2013. It contained both an indorsement and a statement of claim.

12 An amended statement of claim ('ASoC') was filed on 20 June 2014.

13 The writ, the original statement of claim and the ASoC were filed by the plaintiff and were apparently prepared by him. Defamation is a difficult area of the law, even for experienced practitioners and for judges having familiarity with the law of defamation. The difficulty is magnified in the context of the interrelationship between defamation law and the internet. Nonetheless, Google's original application, and the proceeding in this Court, must be considered in the context of the plaintiff's pleaded case.[7] That is subject to one rider. As will appear, the ASoC contains at least one ambiguity. We should approach consideration of this matter on the basis of a reading of the statement of claim most favourable to the plaintiff.

14 We have mentioned that the plaintiff filed an ASoC on 20 June 2014. That was not the entirety of it. In fact, he filed a writ and statement of claim, each of which was in amended form.[8] The amended writ contains this indorsement:

[MILORAD TRKULJA aka MICHAEL TRKULJA Plaintiff endorsement of a statement of claim or of a statement For DEFAMATION and the cause thereof and of the relief or remedy sought in the proceeding JUDGMENT and Aggravated and Punitive DAMAGES, sum $1.000.000.00 and the SUPREME COURT ORDERS That the DEFENDENTS PERMINETLY BLOCK OF PLAINTIFF NAME FROM ALL GOOGLES SEARCH ENGINS in AUSTRALIA and WORLDWIDE TO STOP DISPLYING ARTICLE AND IMAGES OF THE PLAINTFF, s for Images for Melbourne Criminal Underworld Photos and PERMINNETLY BLOCK ALL IMAGES OF/OR FOR MILORAD TRKULJA and MICHAEL TRKULJA. (the Googles Webpage]

15 By his ASoC, the plaintiff pleads, with respect to the defendant, that it is -

(d) A company that intended to publish the defamatory material complained of because any time somebody read that story the first defendant making money, while the systems were automated, those systems were the consequence of computer programs, written by human beings, and were programs were doing exactly what the first defendant ("Google Inc"), and its employees intended and required.

This is, evidently, in part a replication of conclusions expressed by Beach J (as his Honour then was) in Trkulja v Google Inc (No 5), an earlier successful defamation proceeding by the plaintiff against Google.[9]

16 By paragraphs three to 12 of the ASoC, the plaintiff pleads legal propositions with respect to publication of defamatory matter.

The Google Images matter was published on or before 1/12/2012 and 3 March 2014 and current to the persons in the State of Victoria upon those persons accessing the Google site, searching for the plaintiff's names Michael Trkulja and Milorad Trkulja and then viewing and perceiving the matter.

18 To this point, the statement of claim has not identified 'the Google Images matter'. That term is given content by paragraph 16, which reads as follows:

The First and Second Defendant published Defamatory images, of concerning the plaintiff, the matter set out in Annexure A to this Statement of Claim ("the Google Images matter").[10]

19 There follow particulars of publication numbered (i) to (xx). Each particular corresponds with a printout of results produced by the images section of the Google search engine in response to a search term entered by the plaintiff.[11] Those printouts are the 20 pages of Annexure A to the ASoC.[12] For the most part, the printouts are compilations of photographs - so-called 'thumbnails'. The same 20 pages of images, although not in the same order, were Annexure A to the statement of claim in its original form.

20 The date upon which the search was conducted that produced the particular printout is detectable in the case of 19 of the 20 pages. Putting aside the one page of undated images,[13] the earliest search was conducted on 30 September 2012, and the last search on 10 November 2013. Multiple searches were conducted on six days. Since the undated page of images was part of Annexure A to the statement of claim in its original form, it is certain that the search date on the undated occasion preceded 2 December 2013.[14]

21 The predecessor of paragraph 13 of the ASoC[15] commenced as follows: 'The Google Images matter was published on or before 1/12/2012 and 1 December 2013'. That becomes, in the ASoC: 'The Google Images matter was published on or before 1/12/2012 and 3 March 2014 and current'.

22 Reference in each instance to the date 1 December 2012 seems likely to reflect the plaintiff's awareness of the limitation period which is imposed (subject to the operation of s 23B) by s 5(1AAA) of the Limitation of Actions Act 1958. But the operation of those provisions - bearing in mind the fact that five of the image pages were the result of searches conducted before 1 December 2012 - was not the subject of argument, and does not fall for present consideration.

23 It may next be noted that the search terms which resulted in the images being displayed followed a consistent theme. The search terms were as follows:

24 According to the particulars of publication referable to the 'Google Images matter' (for convenience, 'the images matter'), an image of the plaintiff was displayed, 'mixed with' a specified number of 'convicted Melbourne underworld criminals'[16] in response to the search term inputted by the searcher. The number of images of convicted criminals specified as appearing together with the plaintiff's image does not always seem to coincide with the search results forming Annexure A. But be that as may, it was not in contest before us that the images section of the Google search engine displayed, on the searcher's browser, and in response to the search term inputted by the searcher, a compilation of 'thumbnail' images derived from the world wide web as it existed at a moment in time.[17] As depicted on the printouts which form 19 of the 20 pages of Annexure A[18], the thumbnails were mainly reproductions of photographs. We will describe them as 'thumbnails' or 'images' rather than photographs.

25 In each instance of a compilation of thumbnails, the compilation included a thumbnail of the plaintiff. In each instance, also, thumbnails of members, actual or reputed, of the Melbourne underworld appeared. But the trier of fact would immediately notice - and if, by some chance, it was not obvious, then evidence would disclose the fact - that the compilations variously included thumbnails of another, or others, who were not Melbourne underworld figures; and other images altogether. There were thumbnails of a former Chief Commissioner of Victoria Police,[19] two well-known crime reporters,[20] a barrister dressed in wig and gown, a solicitor, a murder victim,[21] a radio and television journalist who is now a senator for the State of Victoria,[22] actors who appeared in film and television productions concerned with the Melbourne underworld,[23] the late Marlon Brando, report headings of defamation proceedings brought by the plaintiff at an earlier time against Yahoo! and Google, the St Kilda pier, and a Melbourne tram. Within the compilations there were also instances of depiction of the Google logo.

26 At time of search, it was possible for the searcher to ascertain the webpage on which any thumbnail had appeared, and thus the context in which it had appeared. That could have been done by 'clicking' on the particular thumbnail. But the plaintiff, in this proceeding, pleads simply that (possibly) the search term itself, and (certainly) the thumbnails constitute the defamatory matter. The search term was the doing of the searcher, and the thumbnails were found on webpages created by other persons. Neither was the work of Google.

27 Had the searcher clicked on a thumbnail, and ascertained the webpage on which the image appeared, the Uniform Resource Locator ('URL') of that webpage, as it existed at that moment in time, could have been ascertained. But the URL of the 'source' webpage could not reliably be determined from the thumbnail as displayed on the particular printout relied upon by the plaintiff.

28 We pause for a moment to say something upon the question whether the plaintiff pleads that the defamatory matter includes the search term as well as the results. The judge below stated in his reasons[24] that the plaintiff alleged that the impugned publication was 'the composite of the search query plus' the images or the web search results, as the case might be.

29 In this Court, there was some debate whether the defamatory matter as pleaded includes the search terms. Senior counsel for Google submitted that the search terms contextualised the alleged defamatory matter, but were not part of the defamatory matter. He submitted also that there was no allegation that Google published the search terms entered by the user. It was as if, in a slander case, a person said something in response to a statement. That is the context in which the meaning of the response would be assessed.

30 We consider that the ASoC is ambiguous in this connection. It is the ambiguity to which we referred at [13] above. So, the plea in respect of the images matter is that Google published 'defamatory images..., the matter set out in Annexure A...'. The particulars to that pleading then identify, in the case of each of the 20 pages, the images and the search term; and the printouts in the annexure show both the search term and the images returned. As we said we would do, we will approach consideration of Google's application[25] on the basis most favourable to the plaintiff. For that reason, despite our reservations, we will assume that the pleaded defamatory matter is the composite of the search term and the search result in each instance.[26]

31 Before referring to the meanings relied upon by the plaintiff, we must identify a second category of defamatory matter pleaded by the ASoC. By paragraph 17 of that document, the plaintiff relevantly pleads that the defendant 'published, of concerning the plaintiff, the matter set out in Annexure B to this Statement of Claim'. This is identified as 'the Google Web matter'. We will call it the 'web matter'.

32 In the original statement of claim, the plaintiff pleaded that the web matter 'was published and was available to be accessed and viewed as a search result ... from 1/12/2012 and 1/12/2013'. In the ASoC, the plaintiff does not plead a date or dates of publication, by contrast with his pleading with respect to the images matter. The ASoC is deficient in this respect, but if that were the only problem confronted by the plaintiff, it would not provide, by any means, a basis for setting aside service.

33 In fact, the dates of alleged publication of the web matter displayed in the documents comprising Annexure B are largely detectable. They span the period 23 November 2012 to 24 June 2013.

34 We should describe the content of Annexure B.[27] Its seven pages fall into different categories.

35 The first page is a screenshot[28] of comments made on a webpage discussing the plaintiff's earlier Google proceeding. The comments, by one 'Picklesworth', mention the plaintiff being 'Streisand'd' (sic), and append an image of autocomplete predictions, the latter associating the plaintiff with the words 'criminal', 'Melbourne crime' and 'underworld'. 'Streisand'd' is a reference to the so-called 'Streisand effect'.[29] The plaintiff's earlier successful proceeding against Google was being said by the comment-maker to have had that effect.

36 The second page of Annexure B is a printout[30] of a search of the terms 'Melbourne-criminal-underworld-figure'. The effect of dashes connecting the terms is that the Google search engine seeks instances of the searched words occurring in sequence. So much was not in debate at the hearing in this Court. But it is still a word search. One result of the search, as shown by the printout, was a number of thumbnails, none of them being of the plaintiff. They were a sample of the images that might have been returned at that time on an image search of the same terms. Another result was a 'snippet' from a webpage dealing with the earlier Google proceeding, and the case which the plaintiff had then raised.

37 The third page is a printout of the results of a search of the words 'Melbourne criminal underworld photos'.[31] One result was a compilation of images, including that of the plaintiff, of the kind referred to in the preceding paragraph. There were also, in that compilation, images of known criminals, a crime reporter, the Google logo and text dealing with the plaintiff's earlier successful defamation proceeding(s). Another result was a 'snippet' from a webpage dealing with a successful defamation proceeding brought by the plaintiff against Yahoo! in 2012.

38 Page four is a printout of the results of a search of the terms 'Melbourne underworld criminals'.[32] The results include the images of four persons including the plaintiff, the other three being well known criminals. This page was specifically mentioned by the judge in his reasons in connection with the issue of defamatory meaning.

39 Pages five, six and seven of Annexure B are of a different character. They purport to be printouts of autocomplete predictions made in the course of Google web searches resulting from input of the prefixes 'michael trk', 'michael trkulj' and 'milorad trkulj'.[33] The predicted auto-completions include 'michael trkulja criminal', 'michael trkulja melbourne crime', 'michael trkulja underworld', 'michael trkulja melbourne underworld crime', 'michael trkulja tony mokbel', and 'milorad trkulja criminal', and 'milorad trkulja tony mokbel'. Other suggested auto-completions are 'michael trkulja v google', 'michael trkulja lawyer', 'Michael trkulja google', 'milorad trkulja google', 'milorad trkulja yahoo', 'milorad trkulja wiki' and 'milorad trkulja lawyer'.

40 We have described pages five, six and seven of Annexure B as purported printouts of autocomplete predictions. Several issues arise. First, page five appears to be a printout of an image located at ozsoapbox.com depicting a Google search in progress with autocomplete suggestions. Second, the affidavits of Michael Herscovici affirmed 18 January 2015 ('the Herscovici affidavit') and Nicholas Weiniger affirmed 3 June 2014 raise serious questions whether pages six and seven of Annexure B, and also page 14 of Annexure A, are in fact what they purport to be. There are said to be strong indications to the contrary. The plaintiff's response to these challenges is presently unknown.

41 Notwithstanding the issues thus raised, Google accepted, both below and in this Court, that the facts needed to be taken at their highest for the plaintiff. For that reason, we take all of the documents in Annexures A and B to be what the plaintiff asserts they are.

42 In respect of the printouts which comprise Annexure B, the plaintiff's claim is that the Google results together with (we will assume) the search terms, without more, are defamatory of him. That is, he eschews recourse to the webpages which underlie pages one to four of Annexure B, and recourse to the webpage which would have been displayed had any of the predicted auto-completions been taken up. Whether any results of further inquiry would have disclosed defamatory matter, or the converse, was thus not revealed at the time when the searches were made; and, because the web is dynamic, and thus constantly changing, it is quite uncertain whether a search made today of any of the predicted auto-completions would disclose the same underlying matter as would have been disclosed in 2012 and 2013 when the searches were conducted.

43 We have referred a number of times to autocomplete predictions. We will explain later what is comprehended by that term.

44 The plaintiff alleges that the images matter and the web matter conveyed imputations which were defamatory of him. He pleads, by paragraph 18 of his ASoC, the following false innuendos:

(a) The plaintiff is a hardened and serious criminal in Melbourne

(b) The plaintiff is a hardened and serious criminal in Melbourne in the same league as convicted murderer Carl Williams, hardened notorious underworld killer Andrew "Benji" Veniamin, hardened and serious and notorious murderer Tony Mokbel and the Mafia Boss Mario Rocco Condello

(c) The plaintiff is an associate of underworld killer Andrew "Benji" Veniamin

(b) (sic) The plaintiff is an associate of Carl Williams Melbourne notorious convicted criminal murderer and drug trafficker;

(e) The plaintiff is an associate of Tony Mokbel, the Australian notorious convicted murderer and drug supplier and trafficker;

(f) The plaintiff is such a significant figure in the Melbourne criminal underworld that events involving him are recorded on a website that chronicles crime in Melbourne criminal underworld

45 Further or alternatively, the plaintiff pleads, with respect to the images matter, the same imputations as matters of true innuendo. The extrinsic facts relied upon are set out in paragraph 20.

46 We should mention also paragraphs 21 to 23 of the ASoC. It is not easy to understand the point which they seek to make. Paragraph 23 refers to reliance upon 'the following extrinsic facts know [sic] to the person to whom the Google Images and web matter was published'. But no extrinsic facts follow. Furthermore, the paragraph is apparently an intended reference to both the images and web matters. For the purposes of considering the present application, we will assume, favourably for the plaintiff, that the extrinsic facts relied upon with respect to the images matter are intended to be relied upon to support the true innuendo pleaded with respect to the web matter by paragraph 19 of the ASoC.

47 We can summarise the plaintiff's pleading thus far as follows: he identifies the allegedly defamatory matter as - (1) compilations of thumbnails returned by the defendant's search engine in response to search terms inputted into the images section of the Google search engine website; (2) search results returned by the search engine in response to search terms inputted into the web section of the website; and (3) autocomplete predictions at different stages of a search of the plaintiff's name. In each instance, we assume, the defamatory matter includes the search term itself. In no instance does the plaintiff rely upon (1) any of the material in the webpages underlying the thumbnails, (2) the material identified in the web search results, or (3) the material which would have been disclosed had any of the autocomplete predictions been taken up.

48 There is another aspect of the ASoC to which we must draw attention.

49 As will be seen, one part of the submissions respecting publication turned upon a written notice which the plaintiff gave the defendant on about 3 December 2012. The notice had two possible areas of relevance: First, as going to the issue of publication. Second, in the context of Part 3 of the Defamation Act 2005 ('the Act').

50 A good deal must later be said about the first possible area of relevance. As to the second, Part 3 is concerned with resolution of civil disputes without litigation. One aspect of such resolution is the ability of the publisher of impugned material to make an offer to make amends in certain circumstances.

51 By s 14(1) of the Act, an offer to make amends cannot be made in two situations. The first, dealt with by subparagraph (a), is if 28 days have elapsed since the publisher was given a concerns notice by the aggrieved person.

52 Sub-section (2) provides for what is required of a concerns notice. Provision is thereafter made for a publisher to request particularisation, for the provision of particularisation, and for the consequences of a failure to provide particularisation.

53 A concerns notice must inform the publisher of the defamatory imputations that the aggrieved person considers are or may be carried by the matter in question.

54 A concerns notice given to the publisher of impugned material is plainly intended to address an allegedly defamatory publication which has taken place; not to deal with defamatory publications which might or might not take place in the future.

55 In the present case, much of the allegedly defamatory matter particularised, and then set out, in Annexures A and B was published, according to those annexures, after the notice was given.

56 Against the areas of (possible) relevance which we have identified, paragraph 26 of the ASoC reads as follows:

1. On or about 3 December 2012, the plaintiff.

(a) Informed the defendant of the nature of the defamatory matter

(b) Demanded that the defendant remove the defamatory images matter from its computer and servers linking or directing internet users to the matter

(c) Requested that the defendant provide details including contact details of the source or sources of the matter: and

(d) Demanded the defendant block the name of Milorad Trkulja and Michael Trkulja from its computers and servers links or directing internet users to the name of "Milorad Trkulja" and "Michael Trkulja"

57 The substance of the actual concerns notice,[34] dated 3 December 2012, reads this way:

3 December 2012

Milorad (aka Michael) Trkulja

...

Email; igoogle@bigpond.com

URGENT

NOT FOR PUBLICATION:

CONFIDENTIAL AND LEGAL TAKE DOWN NOTICE

By Registered Post

Google Australia Pty Ltd

Level 5, 48 Pirrama Road

PYRMONT NSW 2009 AUSTRALIA

AND TO:

By Registered Post

Google Inc

1600 Amphitheatre Parkway

MOUNTAIN VIEW, CA 94043 USA

Concerns Notice: s.14 Defamation Act 2005 (Vic)

Dear Sirs,

My name is MILORAD TRKULJA (also known as (MICHAEL TRKULJA). I am writing with respect to certain articles and photographs, images and text that is accessible by typing into the Google search engine, www.google.com.au 'Melbourne underworld crime photos'. My photographic is published with photographs of Melbourne underworld criminals, drug traffickers and convicted murders. If you click on the link for 'Images for Melbourne criminal underworld photos' the search results shows that my photographic image is mixed with the photographs of Melbourne underworld criminals, drug trafficker and convicted murders. My photograph is accessible by typing into the www.google.com.au/Images (website and images matter).

My Take Down Notice Michael Trkulja and Milorad Trkulja

All defamatory article and images of Michael Trkulja and Milorad Trkulja are accessible from the Google websites www.google.com.au and www.google.com.au/Images containing photographs of "Melbourne underworld criminals". My photographic image is mixed with the Melbourne underworld criminals, drug trafficker and convicted murders and drug baron Tony Mokbal, and his kingpin killer Carl Williams and his 'hit man' Andrew Veniamin Banji. If you click on the Google link for 'Images for Melbourne criminal underworld photos' my photographic image is mixed with the Melbourne underworld criminals, drug traffickers and convicted murders.

My photograph is also accessible by typing into the Google image search engine www.google.com.au/Images. This web page will no doubt confuse people to believe me to be part of the above Melbourne underworld criminals. I enclose for your reference a copy of the defamatory article and photos images of myself ("the Images matter").

In particular, I note that my photograph Michael Trkulja and Milorad Trkulja) are published by the photographs of well-known convicted murders and drug traffickers who are known as 'Melbourne underworld criminals'. (Mafia's Drug Empire)

Take down "Google Autocomplete" of Michael Trkulja and Milorad Trkulja

When you type 'michael trk' into the Google website www.google.com.au and www.google.com the autocomplete results display the following:

1. michael trkulja Melbourne criminal
2. michael trkula
3. michael trkulja underworld
4. michael trkulja Melbourne underworld criminal

h) http://www.google.com.au/#hl=cn&sugexp=les%3Bernk_eprob&gs_rn=0&gs_ri=hp&cn=12&gs_id=le&shr=t&q=michael+trkulja&pf=p&tbo=d&output=search&sclient=nsv-ab&oq=michael+trku&gs_l=&pbx=1&bav=on.2.or.r_gc.r_pw.r_qf&fp=a688cdd845976958&bpel=39314241&blw=1067&bih=486

For your information, I confirm that "Tony" Mokbel (Antonios Sajih "Tony" Mokbel) is a criminal and convicted drug trafficker and murderer from Melbourne Australia. Several years ago Mr Tony Mokbel ago fled from Australia to Europe. In 2008 he was extradited from Greece back to Australia. In July 2012 Mokabel was sentenced to 30 [y]ears in prison for his crimes. My photograph can be viewed by anyone with the appropriate URL address and/or anyone who has previously bookmarked those web pages. As at the date of this letter, the material remains viewable by all Google users WORLD WIDE.

For your information I include in this Take Down Notice, My photograph is placed adjacent to a photograph of worldwide and well known 'Melbourne underworld criminals' and drug traffickers and convicted murders: These are Carl Williams and his Drugs Boss partner Tony Mokbal and they 'hit man' Andrew Veniamin Banji. He was a career criminal and long-time associate of Melbourne underworld criminal Carl Williams and Tony Mokbal

My photograph is placed adjacent to a photograph of Judy Moran convicted Criminal. On the other side of my image is a photograph of a well known underworld criminal and Calabrian Mafia money man the late Mario Condello as well as photographs of the head of the Italian-Australian Mafia, the Alphonse Gangitano. He was a career criminal and long-time associate of Melbourne underworld criminal Graham Kinniburgh. The photographs of these the criminals have been widely circulated in Australia and worldwide.

Users of the Google search engine will see my photograph mixed with those criminals.

As a direct result of the publication of my photograph (of Michael Trkulja aka Milorad Trkula) with images of criminals I have suffered, and will continue to suffer, irreparable injury and harm to my reputation and that of my family in my ethnic community. I have also suffered significant distress, embarrassment and humiliation in my church community.

Accordingly, I demand that by 4.00pm 21 December 2012 Google removes all defamatory articles, text, photographs and images of myself, Michael Trkulja and Milorad Trkulja) from their search results for the search terms, Melbourne Underworld Criminals. and 'Melbourne Criminal Underworld Photos'

I also demand that by 4.00pm 21 December 2012 Google take down or removes the results for their 'Google Autocomplete' the following terms.

1. [http://www.google.com.au/#hl=cn&sugexp=les%3Bernk_ eprob&gs_rn=0&gs_ri=hp&cn=12&gs_id=le&shr=t&q=michael+trkulja&pf=p&tbo=d&output=search&sclient=nsv-ab&oq=michael+trku&gs_l=&pbx=1&bav=on.2.or.r_gc.r_pw.r_qf&fp=a688cdd845976958&bpel=39314241&blw=1067&bih=486](http://www.google.com.au/#hl=cn&sugexp=les%3Bernk_ eprob&gs_rn=0&gs_ri=hp&cn=12&gs_id=le&shr=t&q=michael+trkulja&pf=p&tbo=d&output=search&sclient=nsv-ab&oq=michael+trku&gs_l=&pbx=1&bav=on.2.or.r_gc.r_pw.r_qf&fp=a688cdd845976958&bpel=39314241&blw=1067&bih=486)
2. michael trkulja criminal

(a) michael trkulja Melbourne criminal

(b) michael trkulja underworld

(c) michael trkulja Melbourne underworld criminal

I further demand:

1. That Google provides a suitable offer to pay compensation for the damage to my reputation. Such compensation is to be agreed upon.
2. Pay my legal costs to date.
3. Provide to me the names and email address of the owners/operators of the website: http://ozsoapbox.com/melbourne/a-curious-email-from-michael-trkulja-re-suing-google/ Google has this data because when the Google email was registered, the registrant was provided a link to verify the email.
4. Provide to me the names of owners and/or operators and email addresses in Australia of the website:

http://skombrii.knightsarmy86976.yuku.com Google has this data because when the Google email was registered, the registrant was provided a link to verify the email.

1. Remove the links, http://skombrii.knightsarmy86976.yuku.com from the Google search engines, www.google.com.au and www.google.com.au/images and www.google.com.

Google has this data because when the Google email was registered, the registrant was provided a link to verify the email.

1. Provide to me name of the owner operator of the email: ceco.gakovic@gmail.com. Google has this data because when the Google email was registered, the registrant was provided a link to verify the email.
2. Provide to me the names of owners and/or operators and IP Addresses and email: of http://www.youtube.com/user/PinkP4nter Google has this data because when the Google email was registered, the registrant was provided a link to verify the email.
3. Provide to me the names of owners and/or operators and email addresses in Australia of this email vrackoplacko@gmail.com Google has this data because when the Google email was registered, the registrant was provided a link to verify the email.

For your information, I enclose pages downloaded from 3 http://ozsoapbox/rest-of-australia websites. This page shows the images that are the subject of this Concerns Notice. For You Information I have marked the Defamatory article and images on each pages.

1. Take down 6 websites 10 to 16 below, from the Googles search engines www.google.com.au and www.google.com.au/images and www.google.com
2. http://ozsoapbox.com/rest-of-australia/censorship/trkulja-google-verdict-a-defamation-embarrassment/
3. http://ozsoapbox.com/melbourne/yahoo-held-liable-for-search-results-in-defamation-case/
4. http://ozsoapbox.com/melbourne/a-curious-email-from-michael-trkulja-re-suing-google/
5. http://knightsarmy86976.yuku.com/topic/1908/Re-You-have-questions-Milorad-has-more-answers?page=1
6. http://www.sabotagetimes.com/life/the-australian-criminal-who-won-a-landmark-defamation-case-against-google
7. http://www.youtube.com/watch?v=UwCWgeoXpgO
8. http://www.youtube.com/user/PinkP4nter Copy page URL downloaded from www.google.com.au

Accordingly, I demand that by 4.00pm 21 December 2012 the Google Inc and Google Australia Pty Ltd take down this 6 websites in paragraph 10 to 16 above.

Be advised that Google Australia Pty Ltd and Google Inc is fully responsible for all defamatory images and articles posted on this 6 websites in paragraph 10 to 16 above.

According to the Australia Defamation Law. The Google Inc is the publisher and are responsible for all Google domains published defamatory article and images worldwide. And they are accessible from Australia. I attach the decision in the Supreme Court the landmark deformation case reference Milorad Trkulja v Google Inc LLC & Google Australia Pty Ltd [2012] VSC 533\ also enclosed Judgment by His Honour Justice Beach

http://scv2.webcentral.com.au/judgments/pdfs/T0533.pdf#page=1&navpanes=0&toolbar=1&scrollbar=1&pagemode=none

("the Defamatory imputations")

I am advising you that according to Australian case law a person may be liable not only for direct remarks, but also for conveyed imputations (see Favell v Queensland Newspapers Ltd (2005) 221 ALR). I also remind you that my right to sue you in this jurisdiction of The State of Victoria, Australia, in which I have an established good reputation has been established in the Australian High Court (see Dow Jones and Company Inc v Gutnick (2002) HCA 56).

None of these defamatory meanings are defensible. These imputations are malicious and the comments have damages my good reputation and caused hurt feelings in myself and my family. There was, and is, no factual foundation for any of the imputations. I am not a criminal or "gangster". I remind you that under Australian defamation laws it is incumbent on you, as the defendant, to provide evidence of the truth of this imputation.

As a direct result of the imputation in your publication of the word "gangster", I suffered, and continue to suffer irreparable injury and harm to my reputation and feelings. Moreover I suffered and continue to suffer distress, embarrassment and humiliation within my family, my community and among my church members.

I further demand that you immediately remove and/or take down all articles photographs and images of myself (Michael Trkulja and Milorad Trkulja) referenced in this take down notice that are within your possession, custody or control including from all servers, back-up servers, and other electronic storage devices, remove all links to, and/or information directing viewers to copies of the Defamatory Material.

You would be well advised to comply with these take down notice demands immediately and to inform me in writing of your agreement to do so by 21 December 2012. Please note that in the event that these matters are not agreed, I will instruct my solicitor to issue legal proceedings against you without further notice.

My claim will include aggravated and exemplary damages pursuant to The Defamation Act 2005 (Vic).

This letter does not purport to constitute a complete or exhaustive statement of all of my (Michael Trkulja or Milorad Trkulja) rights, contentions or legal theories. Nothing contained herein is intended neither as, nor should it be deemed to constitute, a waiver or relinquishment of any of My rights or remedies whether legal or equitable, all of which are hereby expressly reserved.

My photos images mixed with "Melbourne underworld criminals" I have suffered, and will continue to suffer, irreparable injury and harm to my reputation. No amount of money can actually compensate me for the distress and shock dealing with My reputation being questioned by my friends and the Community and Church members.

Yours sincerely

Milorad Trkulja (also known as Michael Trkulja)

NB: The Google Australia Pty Ltd and The Google Inc never responded to my Concern Notice dated 1. July 2011.

Please be advised for the court records that I sent a Concerns Notice on July 1st 2011 by registered post to Google Australia Pty Ltd. The registered post number is: AP 512212960015. I also sent a Concerns Notice by registered post to Google Inc on 1 July 2011. The registered post receipt number is RR142488105AU.

On July 22nd 2011, I received an email from removals@google.com signed by 'The Google Team' with an attachment 'Letter from Trkulja to Google- Notice 1/7/2011'. They requested that I send them the same publication that I already had sent with the landing page URL. A copy of the email will be produced to the Supreme Court.

The perfect example in December 2007, in the Supreme Court Melbourne Australia:

Plaintiff's Mark Forytarz & Paul Castran v Google Australia Pty Ltd. The Google Australia Pty Ltd has removed defamatory publications from www.google.com.au and www.google.com.au/images

Second example Dr Duffy who suing v Google Australia Pty Ltd and Google Inc in Adelaide South Australia.

Google Australia Pty Ltd has removed part of defamatory publication from www.google.com.au and www.goole.com.au/images 10 day before Dr Duffy served Concerning Notice to Google Inc in USA..

The Google Australia Pty Ltd did not request that Dr Duffy provide them with Urls.

58 The demands made by sub-paragraphs (b), in part, and (d) of paragraph 26 of the ASoC fit in with paragraph C of the prayer for relief, which reads as follows:

C. A permanent injunction by the Supreme Court orders that the defendants permanently block Google Images and web searches of the Plaintiff's names "Milorad Trkulja" and "Michael Trkulja" from its computers and servers and remove all links from its computers and servers linking to the Google webs and images users from Australia.

59 At least in part, what was purportedly a concerns notice was outside the statutory framework. Nonetheless, the notice in some part pertained to the allegedly defamatory matter relied upon in the ASoC; and it was said for the plaintiff to bear upon the question whether the defendant published the impugned matter.

60 Google replied at length to the plaintiff's notice on 16 January 2013. The reply is exhibit JHC-9 to the affidavit of James Clark sworn 30 January 2015 and filed on behalf of the defendant. Google stated that it had removed the URL of one 'ozsoapbox' webpage from future search results; and, specifically without admission, that it had blocked certain predictions and queries respectively relating to the plaintiff from appearing as part of the autocomplete and related searches features on google.com.au. Only if this proceeding was to continue might the sufficiency or otherwise of Google's response fall for consideration. The same observation applies to subsequent correspondence between the plaintiff and Google.

The plaintiff's earlier defamation proceeding against Google

61 In rejecting Google's submission that it could not be held to be the publisher of the impugned matter, the judge relied, as will be seen, upon a ruling of Beach J in the earlier Google proceeding. That ruling was made on a non obstante veredicto motion by Google, the jury having found by answers to questions that Google had published the matter upon which the plaintiff relied.

62 Apart from concluding that the reasons of Beach J supported a conclusion that there was publication of the impugned matter in the present case, the judge gave considerable attention to whether those reasons gave rise to an issue estoppel upon that question. His Honour did not, however, finally decide the point, although he stated that the possible application of such an estoppel - depending upon it being pleaded in response to Google's defence - was another reason why it could not be concluded that the plaintiff's claim had no real prospects of success.

63 It is desirable, in the event, to say something immediately about the circumstances of the earlier proceeding. There, the plaintiff relied in part upon the results of a Google image search of the term 'michael trkulja'. The search returned 3810 results. Amongst the first 20 results were an image of the plaintiff, two Facebook images of another man (not identified as a criminal), and two images of persons associated with Melbourne crime. Each of the five images was captioned 'Michael Trkulja'. The images were 'thumbnails'.

64 The image results were quite different in form to the results relied upon in the present proceeding. They were captioned, and in each instance the source webpage was identified by a truncated part of the domain name. Contrast the situation described at [26] above.

65 Within the image search aspect of his claim, the plaintiff relied, too, upon the content of an underlying webpage, accessed by clicking on a thumbnail of his face in two instances. That page largely consisted of an online article on the site 'Melbourne crime'. It dealt with the plaintiff having been shot in 2004, of assertions which he made in 2007 that he knew the attacker's identity, and with a request that police re-open their investigation of the incident.

66 The plaintiff also relied upon the results of a Google web search of the term 'michael trkulja'. Depicted on a printout were the first ten of about 185 000 results, one of which was a snippet under the headline 'michael trkulja - Melbourne crime - underworld - ganglands', together with a hyperlink to a particular webpage. The snippet referred to the plaintiff having been shot in 2004. The plaintiff relied also upon the content of the webpage identified by the snippet. The searcher was taken to the same article as that to which we referred a moment ago.

67 Thus, in the case of both the image search and the web search, the plaintiff relied not simply upon the thumbnails or snippet (as the case might be), but also upon the content of an underlying webpage.

68 The plaintiff pleaded a number of imputations - both false and true innuendos. He succeeded upon only one imputation - that is, that he was 'so involved with crime in Melbourne that his rivals had hired a hitman to murder him'. It could not be doubted that this imputation mainly rested on the content of the online article to which we have referred.

69 We do not understand that there ever was, or is now, any evidence that the plaintiff had, or has, any connection with any of the actual or reputed criminals depicted in the various Google image and web search returns. Rather, it appears, the whole edifice underpinning the defamation which is now pleaded traces back to the fact that the plaintiff was shot by an unknown assailant[35] in 2004, and to the incident being mentioned in an online 'crime' webpage. Building upon those matters and his earlier successful defamation proceedings against Google and Yahoo! (which themselves sparked both newspaper reporting and online reporting and comment), it appears that repeated searches over the ensuing years by the plaintiff (and evidently others, some of whom are certainly in his 'camp') of terms of the kind set out at [23] above have produced a self-perpetuating series of responses.

70 It is unnecessary to recapitulate every part of the judge's reasons. We focus upon the matters which were decisive.

71 His Honour noted that Google's 'primary contention' was that a search engine proprietor could not be a publisher, either before or after receiving notice of any allegedly defamatory publication, in respect of the results of a search enquiry. Google's second contention was that the plaintiff had no real prospect of success in establishing that the search engine results underpinning his claim were defamatory. Its third contention was that as a search engine operator it should have the benefit of an immunity from liability for defamation. His Honour ultimately concluded that:

The submissions advanced on behalf of Google fall well short of establishing that Mr Trkulja has no real prospect of establishing at trial that Google is a publisher and/or that any of the material about which he complains is defamatory. Google's invitation to the Court to confer an immunity out of thin air is rejected.[36]

72 We must say more about the reasoning which led to those conclusions.

73 Respecting the publication issue, the judge stated that Beach J had been plainly correct in observing, in the earlier Google proceeding, that '[t]he question of whether a particular internet service provider might be a publisher in respect of defamatory material published through or via, or with the assistance of a particular internet product is "fact sensitive"',[37] this observation having been endorsed by Deputy Justice Ng of the Hong Kong High Court in Yeung v Google Inc.[38]

74 That led on to his Honour identifying the factual circumstances relied upon by the plaintiff. There were two bodies of material - the images and web matter set out in Annexures A and B.[39]

75 Having defined the defamatory meanings in respect of which, in reliance upon false innuendo, the plaintiff relied, the judge then stated that the plaintiff relied upon the composites of the search terms and the results. We have already said that, despite ambiguity, we will treat the pleading as being to that effect.

76 The judge then identified the plaintiff's complaint more closely:

[H]e complains about Google's search engines linking his name and/or image with material relating to the Melbourne criminal underworld in the results that Google's search engine generates. This is illustrated by the fourth page of Annexure B to the amended statement of claim. Under the heading 'Melbourne underworld criminals' there are four separate images: Mr Trkulja, Judith Moran, Matthew Johnson and Tony Mokbel. Mr Trkulja's complaint is that the search term 'Images for Melbourne underworld criminals' has generated his image linked with three convicted murderers.[40]

77 As we earlier observed, the page selected by the judge as illustrating the nature of the plaintiff's claim - page four of Annexure B - was the only page of either Annexure in which a thumbnail of the plaintiff appeared exclusively in the company of thumbnails of convicted criminals. Moreover, accurate as his Honour's part reference to the particular page was, it did not address the question whether it was permissible for the plaintiff to rely upon that page alone (but considered in its entirety) to establish potentially defamatory meaning.

78 His Honour next stated that the plaintiff did not allege that Google was 'a subsidiary or secondary publisher of defamatory matter which had been published by a third party'.[41] This led on to his observation that, 'If there is any author of the material that is the subject of Mr Trkulja's complaints, it can only be Google. Either Google is the publisher of the material complained of or there is no publisher at all'.[42]

79 We pause again to make these observations. First, as will be seen, a question arose in the course of the hearing in this Court as to the possible relevance of the notice given by the plaintiff to the defendant, dated 3 December 2012. The debate led into the question whether the plaintiff advanced a case other than that the defendant was a primary publisher. The ultimate answer to that question was 'no'.

80 Second, in the observation noted at [78] above, the judge must be understood to have used the word 'publisher' in its legal sense in the context of defamation law. There is no doubt that, in response to terms inputted by the plaintiff, the Google search engine produced search results, and made autocomplete predictions, which were visible on the web browser of the searcher, and which could be, and were, printed out by the searcher. The question which must presently be decided, however, is whether the defendant has established that the plaintiff has no real prospect of making out his plea that those search results, and/or the autocomplete predictions, were published by it in the legal sense.

81 The judge next noted submissions for Google that in no circumstances, whether on notice or not, could it be, as proprietor of a search engine, the publisher of search results produced automatically by algorithms in response to a user's request. He noted Google's reliance on Bunt v Tilley,[43] Metropolitan International Schools Ltd v Designtechnica Corporation,[44] Tamiz v Google Inc,[45] and Bleyer v Google Inc.[46] Google's submission, he noted, was that '[t]he principled approach is to focus on the defendant's intention to publish the actual matter complained of'. Here, no such intention was present, even as a matter of inference.

82 Foreshadowing his ultimate conclusion, the judge then said this:

Assuming in Google's favour that it had no intention of publishing the particular combination of words and images which are the subject of Mr Trkulja's allegations, it does not follow that Mr Trkulja has no real prospect of establishing at trial that Google is a publisher of the alleged defamatory material. To the contrary, provided that Google had an intention to publish the results that its search engine produced and of which Mr Trkulja complains, Google may still be found to be a publisher.[47]

83 The last sentence of the passage just cited was at the heart of the reasoning of Beach J in Trkulja (No 5). So it is unsurprising that the judge in the present case then proceeded to analyse the circumstances and the reasoning in that earlier matter. It had been submitted for Google that the ruling of Beach J was clearly wrong. Having analysed the ruling, his Honour rejected the submission. He did so in part by consideration of the reasons of Beach J, in part by his own analysis of the authorities upon which Google relied, and in part in reliance upon authorities which, in his Honour's view, supported the conclusion reached by Beach J.

84 It is not necessary to set out in any detail the judge's reasons for concluding that the authorities relied upon by Google did not impugn the conclusion reached by Beach J. We do mention, however, the judge's conclusions that - (1) the Court of Appeal decision in Tamiz supported the position that, once placed on notice by the plaintiff of the alleged defamatory material, it was arguable, by its failure to remove the material within a reasonable period, that Google made itself responsible for the continued presence of the material and so became a publisher of that material, if it was not a publisher from the outset;[48] and (2) that the decision of McCallum J in Bleyer did not support Google's contention that in no circumstance, prior to or post-notification, could it be a publisher. McCallum J had expressly refrained from making a finding that, post-notification, Google could not be a publisher.[49]

85 Further, the judge stated that he disagreed with the conclusion of McCallum J that Beach J had erred in Trkulja (No 5) in concluding that '[t]he performance of the function of the algorithm in that circumstance is capable of establishing liability as a publisher at common law'. Rather, Webb v Bloch,[50] Thompson v Australian Capital Television Pty Ltd,[51] Oriental Press Group v Fevaworks Solutions Pty Ltd[52] and Duffy v Google Inc[53] supported the proposition that the question whether Google is a publisher is to be determined by reference to its participation in the publication.

86 We should refer also to the following passages in the judge's reasons, which endorsed the correctness, as his Honour concluded, of the ruling of Beach J in Trkulja (No 5):

1. Employees of Google create the algorithms which automatically generate the search results which are the subject of Mr Trkulja's complaint in the current proceedings. Arguably, they stand in a position similar to the solicitor Norman in Webb. Their skill and expertise is employed by Google for the purpose of creating a search engine which publishes results to online users. As Beach J observed in Trkulja, the automated systems produced the search results which they were designed to produce. If Google intends its search engines to publish material on the internet in response to user queries, whatever that material might be, it is arguable that this makes Google a publisher of the material.
2. In [18] and [19] of Trkulja, Beach J drew an analogy between the position of Google as the operator of the search engine and authorities which have held newsagents and libraries as being publishers notwithstanding the absence of a specific intention to publish defamatory material. Plainly, Beach J was referring to a very long line of authority dating back to the late 19th century to the effect that newsagents and libraries may be publisher notwithstanding the absence of an intention to publish particular material which is defamatory. This line of authority commences with Emmens v Pottle and Vizetelly v Mudie's Select Library.
3. No Australian court has held that Emmens or Vizetelly were wrongly decided. Both cases were cited with approval by the High Court in Thompson. In Thompson, Channel 7 operated a television station in the Australian Capital Territory. On 21 February 1984, it broadcast a program entitled 'The Today Show'. The program was produced live by Channel 9 from its studios in Sydney. The program included a live interview with a young woman during which it was alleged that her father, Mr Thompson, had committed incest with her from the time she was seven years old and had fathered a child to whom she had given birth when she was 14. There was no evidence to indicate that there was any truth in this allegation. Thompson sued Channel 7 seeking damages for defamation. The High Court confirmed the availability under Australian law of the defence of innocent dissemination. Brennan CJ, Dawson and Toohey JJ cited with approval the following passage from Duncan and Neill on Defamation:

It is submitted that it would be more accurate to say that any disseminator of a libel publishes the libel but, if he can establish the defence of innocent dissemination, he will not be responsible for that publication.[54]

87 The judge, as will now be evident, repeatedly returned to the question of intention, characterised as the intended operation of the Google search engine to produce the results which were displayed. A further example of his Honour's reasoning in this connection was his statement that:

Google authorises for its own commercial purposes the publication of its search engine's results on the internet. Its search engine was established and is maintained by its employees. The results generated by its search engines are the result of the operation of the algorithms created by its employees. Even if Google does not intend its search engine to produce results linking Mr Trkulja's image with those of convicted criminals, there is no suggestion that this outcome was anything other than the product of the intended operation of the search engine. Applying the reasoning of the High Court in Thompson, Google's ignorance of the specific material generated in response to any given user query does not prevent it from being found liable as a publisher.[55]

88 The last sentence of the paragraph just cited was an apparent reference to the situation whereby a broadcaster of electronic material received from another might be accounted a primary publisher. But whether that was the foundation for his Honour's conclusion with respect to publication in the instant case is uncertain. Of this, more later.

89 The judge's understanding was that Google, by its submissions, was seeking to conflate two discrete issues - (1) whether it had published defamatory material; and (2) whether it could avail itself of the defence of innocent dissemination. He surmised, perhaps, that Google would conduct a case that it was not a publisher at all, but that if it was held to be a publisher, then it was a secondary publisher and could avail itself of an innocent dissemination defence.

90 In the end result, as we noted earlier, his Honour rejected Google's argument that it could not be held to be a publisher of the allegedly defamatory material in the circumstances of this case. Indeed, he concluded that it was

strongly arguable that Google's intentional participation in the communication of the allegedly defamatory results relating to Mr Trkulja to users of the search engine supports a finding that Google published the allegedly defamatory results.[56]

91 It remains to mention one other aspect of his Honour's reasoning on the publication issue. On the footing that the ruling in Trkulja (No 5) was applicable to the circumstances, albeit different, in the present case, and having regard to his conclusion that the ruling was correct, his Honour gave consideration to whether an issue estoppel might arise whereby Google could be precluded from arguing a 'no publication' point. He did not finally resolve that issue, but he did state that it was arguable that the plaintiff 'may have grounds for contending that the reasoning of Beach J in Trkulja (No 5) at [18] and [28] was legally indispensable to his judgment'.[57] Whilst he would have concluded in any event that Google had not established that the plaintiff's 'publication allegations have no real prospects of success... Nevertheless, the potential availability of a plea of issue estoppel reinforces this conclusion.'[58]

92 We turn to the second contention advanced below, that is, that the impugned material was not defamatory. The judge dealt with the contention quite shortly:

1. For present purposes, it is necessary for Google to establish that Mr Trkulja has no real prospect of establishing that the images and completions generated by Google's search engine carry the alleged defamatory imputations for which he contends. Whilst Mr Trkulja may ultimately be unsuccessful at trial in establishing that the material in question is defamatory, there is no proper basis for concluding that he has no real prospect of establishing that any of the material he complains of is defamatory.
2. I have referred earlier in this judgment to a compilation of images of Mr Trkulja alongside images of convicted criminals Judith Moran, Matthew Johnson and Tony Mokbel. These images form part of Annexure B to Mr Trkulja's statement of claim and appear under the heading 'Images for Melbourne Underworld Criminals.' It is certainly arguable that a reasonable internet search engine user would look at this compilation of images and assume that Mr Trkulja was also a convicted criminal.[59]

93 There, as earlier in his reasons, his Honour referred to the only page of the 27 pages comprised within Annexures A and B which displayed (together with other content) a compilation of images of the plaintiff and other persons, each of the others being a convicted criminal. The judge evidently attributed particular significance to this compilation.

94 Respecting the immunity argument advanced by Google, the judge concluded that '[i]f Google is to have immunity from suit, it must be bestowed upon it by the legislature.'[60] The Act prescribed an entire division devoted to statutory defences to defamation actions. So, his Honour concluded, 'Parliament's willingness to prescribe these statutory defences militates heavily against the introduction of an internet search engine immunity into the common law of Australia.'[61]

1. The learned primary judge erred in failing to find that the plaintiff had no real prospect of establishing that Google Inc is a publisher of the materials upon which he sues.
2. The learned primary judge erred in finding that it was arguable Google Inc is a publisher of automatically generated (i) search results and (ii) completions of search queries, because Google Inc intentionally developed and made available to the public its search engine.
3. The learned primary judge erred in failing to find that, for a defendant to be held liable as a 'publisher', the plaintiff must prove intention to communicate the words in question.
4. The learned primary judge erred in failing to have regard to the evidence in this case, which showed that:

1. the web is comprised of many trillions of documents (webpages), the contents of which vary enormously; 2. it is easy for anyone to upload materials on the web;

1. the web is continually changing and growing, and as a consequence the same search performed at different times can (likely, will) return different results;

1. the same search performed by two different users can (likely, will) return different results even if the two users should make that search at exactly the same time; 2. in order for the Google search engine to be of any use in locating information that is on the web at any given time, it must be designed to utilise fully automated programs that must be capable of being executed extremely quickly; 3. the Google search engine only returns such results to a user in response to that user's query; 4. every month, over 100 billion searches are made by users of the Google search engine and, of those searches, 15% (or approximately 500 million each day) are ones that have never been made before; and 5. the Google search engine is not capable of ascertaining what meanings may be conveyed by the words or images constituting the content of a particular webpage, nor is the Google search engine capable of ascertaining meanings in respect of trillions of documents and/or the unfathomably large number of different possible ways in which parts of those documents may be combined in the search results returned to a user.
2. The learned primary judge erred in failing to find that intention with respect to the words in question is not established by showing no more than a defendant developed and made available to the public a mechanism capable of returning, out of an unfathomably large number of possible search results, the ones complained of.
3. The learned primary judge erred in finding that Google Inc had wrongly sought to conflate two issues, namely whether it had published defamatory materials and, if so, whether it could avail itself of a defence of innocent dissemination.
4. The learned primary judge erred by conflating the two issues identified in ground (6) above, and by wrongly considering that the defence of innocent dissemination was relevant to whether the plaintiff could discharge the onus of proving that Google Inc was the 'publisher' of automatically generated (i) search results and (ii) completions of search queries.
5. To the extent that the learned primary judge relied upon 'the potential availability of a plea in issue estoppel' to support the conclusion that Google Inc had failed to show that the plaintiff had no real prospects of success, the learned trial judge erred in so doing, for the following reasons:

1. identification of the fact and extent of any issue estoppel depends solely on the earlier decision said to give rise to the estoppel - it does not require consideration of what may be pleaded in reply; 2. for Google Inc to be precluded by issue estoppel there had to have been, in the first Trkulja decision, a determination that it was the publisher of all search results that may be returned by its search engine, without which determination the earlier decision could not stand;

1. there was no such determination made in the first Trkulja decision;

1. further, no such determination even if it were made (which it was not) was legally indispensable to the dismissal of Google Inc's non obstante application.
2. The learned primary judge erred in failing to find that the plaintiff had no real prospect of establishing that automatically generated (i) search results and (ii) completions of search queries were defamatory of him.
3. The learned primary judge, having decided he would consider and follow the decision of Blue J in Duffy v Google Inc (handed down after judgment had been reserved), erred in failing to:
   1. follow that decision, insofar as it holds that automatically generated completions of search queries are not capable of being defamatory; and
   2. apply that decision, by parity of reasoning, to automatically generated search results consisting of combinations of images.
4. The learned primary judge erred in failing to find that Google Inc, as the owner/operator of a search engine, is immune from liability in defamation when the matters complained of are automatically generated (i) search results and (ii) completions of search queries.

96 We make two observations. First, in substance, the 11 grounds raise again the three submissions advanced by Google below - that is, (1) Google could not be held to be a publisher of the matter complained of (grounds 1 to 8); (2) the plaintiff had no prospect of proving that the impugned matter was defamatory of him (grounds 9 and 10); and (3) Google ought be immune from liability in defamation in respect of automatically-generated search results and autocomplete predictions produced in response to search enquiries (ground 11).

97 Second, the grounds allege many errors on the judge's part. But the subject of the application for leave to appeal, and of the appeal if leave be granted, is the orders which his Honour made, most particularly the order dismissing Google's application by its further amended summons of 12 February 2015. Discerned errors in reasoning would not necessarily impugn the correctness of the orders made.

98 The question is whether Google has established that the plaintiff has no real prospect of making out his plea that Google published the impugned material. In order to answer that question, we think it is logical to consider the concept of publication in the context of defamation principles as developed by the common law - at first with respect to print, later with respect to radio and television broadcasts; next to consider the impact, if any, of the Act upon that body of law; and then to consider authorities touching upon the operation of defamation law in the context of the internet.

99 It must immediately be said, in our opinion, that to speak of the operation of defamation principles in the context of the internet is an oversimplification which is apt to mislead. There are a number of different ways, having distinct purposes, in which material can be placed on, and be accessible via, the internet. Authorities to which we will refer demonstrate the point. It does not follow, because the law of defamation has been held to operate in a particular way with respect to certain material placed on and accessible via the internet, that the same result should obtain with respect to material having a different genesis.

100 In a well-known passage in Webb,[62] a case concerning defamation by a written document, Isaacs J approved the following statement as to the level of participation necessary to constitute a person a publisher:

In Parkes v Prescott Giffard QC quotes from the second edition of Starkie:

All who are in any degree accessory to the publication of a libel, and by any means whatever conduce to the publication, are to be considered as principals in the act of publication: thus, if one suggest illegal matter in order that another may write or print it, and that a third may publish it, all are equally amenable for the act of publication when it has been so effected.

101 So it is, in the case of a newspaper article, that not simply the author, but also the editor, the publisher and the printer[63] will be the publishers of defamatory matter; whilst in the case of a television broadcast, the 'on air' author of a defamatory statement, the producer of the programme and the broadcaster will be legally responsible.[64]

102 In particular circumstances, there also fall into this class of publishers persons who played no part in the chain of distribution, but who are caught by the act of another. Instances are the situations which arose in Byrne v Deane[65] and Urbanchich v Drummoyne Municipal Council.[66]

103 In Byrne, the plaintiff was a member of a golf club. The two defendants were the proprietors of the club, and the female defendant was also its secretary. Several automatic gambling machines had been kept by the defendants on the club premises for the use of members. Someone gave information to the police, and the machines were removed. On the following day, a document which was variously described in the Court of Appeal as a 'lampoon' and 'doggerel verse' was posted on the wall of the club. The plaintiff asserted that it defamed him, by implying that it was he who had got the police involved. The document's existence was brought to the attention of the defendants, but they did not remove it. The plaintiff succeeded at trial, but failed on appeal, a majority holding that the words complained of were not capable of a defamatory meaning.

104 For present purposes, all that is relevant is what their Lordships said about publication.

In my judgment the two proprietors of this establishment by allowing the defamatory statement, if it be defamatory, to rest upon their wall and not to remove it, with the knowledge that they must have had that by not removing it would be read by people to whom it would convey such meaning as it had, were taking part in the publication of it.[67]

106 Lord Justice Slesser was of opinion that the plaintiff had failed to show publication against the male defendant, but that there had been some evidence of publication on the part of the female defendant. He said this:

I think having read it, and having dominion over the walls of the club as far as the posting of notices was concerned, it could properly be said that there was some evidence that she did promote and associate herself with the continuance of the publication in the circumstances after the date when she knew that the publication had been made.[68]

107 Lord Justice Greene, having concluded that the defendants had ample power to remove the notice, said this:

It is said that as a general proposition where the act of a person alleged to have published a libel has not been any positive act, but has merely been the refraining from doing some act, he cannot be guilty of publication. I am quite unable to accept any such general proposition. It may very well be that in some circumstances a person, by refraining from removing or obliterating the defamatory matter is not committing any publication at all. In other circumstances he may be doing so. The test it appears to me is this: having regard to all the facts of the case is the proper inference that by not removing the defamatory matter the defendant really made himself responsible for its continued presence in the place where it had been put?[69]

108 His Lordship held, in the circumstances, that both defendants ought be regarded as being consenting parties to the continued presence of the notice.

109 In Urbanchich, posters with a photograph of a number of persons in Nazi uniforms, in the company of Hitler, were glued to bus shelters under the control of the Urban Transit Authority of New South Wales ('the Authority'). One of the faces in the photograph was circled. Words on the poster identified that person as the plaintiff and stated that an extreme group of which he was a member was establishing itself in a local council. The plaintiff drew the Authority's attention to the posters and asked that they be removed. They remained in place for another month.

110 The plaintiff brought an action for defamation against the council and also the Authority. There was a separate trial of the question whether, in effect, there was evidence fit to go to trial of publication by the Authority.

111 Justice Hunt ruled, on the facts presented for his consideration, that the Authority's conduct was capable of amounting to publication of the posters. His Honour made these statements of principle:

In a case where the plaintiff seeks to make the defendant responsible for the publication of someone else's defamatory statement which is physically attached to the defendant's property, he must establish more than mere knowledge on the part of the defendant of the existence of that statement and the opportunity to remove it. According to the authorities, the plaintiff must establish that the defendant consented to, or approved of, or adopted, or promoted, or in some way ratified, the continued presence of that statement on his property so that persons other than the plaintiff may continue to read it - in other words, the plaintiff must establish in one way or another an acceptance by the defendant of a responsibility for the continued publication of that statement.

Such conduct on the part of the defendant may of course be established by inference. Indeed, in most cases there will be no evidence of any such acceptance by the defendant expressly, and it can only be established by inference.[70]

The proposition that conduct of a passive nature cannot amount to publication, which is also asserted in the majority judgment in Wolfson v Syracuse Newspapers Inc (at 642), was decisively rejected in Byrne v Deane at 837. Greene LJ said that he was quite unable to accept any such proposition. Nor am I. There is nothing in Byrne v Deane to suggest that the decision turned upon some direct relationship between the occupier and the persons who used the premises which gave rise to a duty on the part of the occupier to remove the defamatory material.

...

There will, of course, always be issues (which will usually be for the jury to determine) as to whether the defendant had the ability to remove the defamatory statement and whether the time given before the commencement of the action was a reasonable one in which to do so.[71]

112 It is to be noted that, at common law, a defendant who was 'caught' by the application of the two authorities which we have just discussed became a primary publisher of defamatory material only from the date when that defendant was taken to have authorised or acquiesced in the earlier publication of that material by another.

113 The situation just mentioned was conceptually different to that which applied in certain cases where a defendant intendedly, but unwittingly, distributed material within which defamatory matter appeared. In some instances, such a person - described as a 'secondary' or 'subordinate' publisher - was absolved from liability for publication.

114 The starting point was the decision of the English Court of Appeal in Emmens v Pottle,[72] where the plaintiff was defamed in a newspaper which was sold by the defendants in the ordinary course of their business. A jury held that neither of the defendants knew that the newspapers contained libels, that it was not by negligence on their part that they did not know there was any libel, and further, that they did not know that the newspaper was of such a character that it was likely to contain libellous matter, nor ought they to have known so.

115 The judge having entered judgment for the defendants, the plaintiff unsuccessfully appealed. Lord Esher MR said this:

I do not intend to lay down any general rule as to what will absolve from liability for a libel persons who stand in the position of these defendants. But it is a material element in their position that the jury have found in their favour as they have done. I agree that the defendants are prima facie liable. They have handed to other people a newspaper in which there is a libel on the plaintiff. I am inclined to think that this called upon the defendants to show some circumstances which absolve them from liability, not by way of privilege, but facts which show that they did not publish the libel. We must consider what the position of the defendants was. The proprietor of a newspaper, who publishes the paper by his servants, is the publisher of it, and he is liable for the acts of his servants. The printer of the paper prints it by his servants, and therefore he is liable for a libel contained in it. But the defendants did not compose the libel on the plaintiff, they did not write it or print it; they only disseminated that which contained the libel. The question is whether, as such disseminators, they published the libel? If they had known what was in the paper, whether they were paid for circulating it or not, they would have published the libel, and would have been liable for so doing. That, I think, cannot be doubted. But here, upon the findings of the jury, we must take it that the defendants did not know that the paper contained a libel. I am not prepared to say that it would be sufficient for them to show that they did not know of the particular libel. But the findings of the jury make it clear that the defendants did not publish the libel. Taking the view of the jury to be right, that the defendants did not know that the paper was likely to contain a libel, and, still more, that they ought not to have known this, which must mean, that they ought not to have known it, having used reasonable care - the case is reduced to this, that the defendants were innocent disseminators of a thing which they were not bound to know was likely to contain a libel. That being so, I think the defendants are not liable for the libel.[73]

116 Lord Justice Cotton concurred. Lord Justice Bowen also agreed in the result, but sounded a note of caution:

It seems to me that the defendants are no more liable than any other innocent carrier of an article which he has no reason to suppose likely to be dangerous. But I by no means intend to say that the vendor of a newspaper will not be responsible for a libel contained in it, if he knows, or ought to know, that the paper is one which is likely to contain a libel.

117 It is to be observed that Lord Esher MR concluded, the defendants having satisfied the burden of showing circumstances which absolved them from liability, that they did not publish the libel.

118 Then followed Vizetelly v Mudie's Select Library Ltd.[74] There, the proprietors of a circulating library lent out and sold copies of a book which, unknown to them, contained a libel on the plaintiff. A jury concluded, on the evidence, that the defendants had not used due care in the management of their business. Had they done so, they would have discovered the existence of the libel. The jury having returned a verdict for the plaintiff, the defendants unsuccessfully appealed against judgment.

The judgment which I am about to give is based upon the special facts of the particular case, and I do not wish to be considered as laying down any general rule of law applicable to all circulating libraries. In this case I think that there were circumstances which justified the jury in saying that the defendants published the libel of which the plaintiff complains. The defendants having lent and sold copies of the book containing that libel, prima facie they published it. What defence, then, have they? None, unless they can bring themselves within the doctrine of Emmens v Pottle.[75]

120 His Lordship concluded that the state of the evidence was such that the jury was entitled to conclude that it was through negligence on the part of the defendants that they did not find out that the book libelled the plaintiff. Thus, his Lordship said:

They failed to do what the defendants in Emmens v Pottle succeeded in doing, namely, prove that they did not publish the libel.

121 Lord Justice Vaughan Williams agreed. He observed that:

What I understand Emmens v Pottle really to decide is that the innocent publication of defamatory matter, i.e., its publication under such circumstances as rebut the presumption of any malice, is not a publication within the meaning of the law of libel.

122 Lord Justice Romer likewise agreed in the result. Like AL Smith and Vaughan Williams LJJ, his Lordship treated proof of blamelessness by the defendants as resulting in there having been no publication by them.[76]

123 Before the commencement of the Act, there was debate whether it was correct to analyse circumstances such as we have been discussing as involving publication subject to a defence of innocent dissemination; or rather, whether proof of innocent dissemination led to the conclusion that there had been no publication by the particular defendant. That debate was still alive when the High Court decided Thompson.[77]

124 Thompson concerned the liability or otherwise of a television station which broadcast to viewers in the Australian Capital Territory a current affairs programme produced by another station - the broadcast being, for practical purposes, contemporaneous with the receipt of the content from its producer. The content of the broadcast defamed the plaintiff. The broadcaster had taken no part in the production of the programme, but it had the ability to control and supervise the material it televised. It was the broadcaster's voluntary decision to broadcast live. One argument pursued by the broadcaster was that it was merely a subordinate disseminator and had innocently disseminated the material.

125 In their joint judgment, Brennan CJ, Dawson and Toohey JJ discussed the 'no publication' analysis in Emmens and Vizetelly.[78] Their Honours then said that:

Despite its somewhat muddied origins, the defence of innocent dissemination has become well accepted in the United Kingdom. In Australia Emmens v Pottle has the somewhat reluctant endorsement of the New South Wales Court of Appeal in McPhersons Ltd v Hickie.[79]

126 In resolving the instant matter, their Honours said this:

There is no reason in principle why a mere distributor of electronic material should not be able to rely upon the defence of innocent dissemination if the circumstances so permit.[80]

It is true that Channel 7 did not participate in the production of the original material constituting the program. But Channel 7 had the ability to control and supervise the material it televised. Channel 7's answer is that time did not permit monitoring the content of the program between its receipt at Black Mountain and its telecast from the studios in the Australian Capital Territory. That may well be so but it by no means follows that Channel 7 was merely a conduit for the program and hence a subordinate disseminator. It was Channel 7's decision that the telecast should be near instantaneous, a decision which was understandable given the nature and title of the program but which was still its decision.

Without, at this point, trespassing into the second question, namely, whether the defence of innocent dissemination was made out, the nature of a live to air current affairs program carries a high risk of defamatory statements being made. In those circumstances it would be curious if Channel 7 could claim to be a subordinate disseminator because it adopted the immediacy of the program. It did that for its own purposes, that is, to telecast to viewers in the Australian Capital Territory and adjoining areas. The agreement it made with PBL did not, as we have already observed, oblige Channel 7 to broadcast the program. Pursuant to the agreement Channel 7 acquired a licence "to broadcast by television transmission only" episodes of a current affairs program, "The Today Show". It broadcast for its own purposes, not as agent for PBL or Channel 9, and it "authorised" the broadcast in any meaningful sense that term has.[81]

127 Justice Gaudron identified the questions for decision this way:

In terms of the usual formulation of the rule, two questions are raised by the claim that Channel 7 is entitled to raise a claim that it innocently disseminated the material of which the appellant complains. The first is whether a broadcaster who retransmits televised material necessarily knows or ought to know the character of the material which it retransmits so that, as a matter of fact, the rule cannot apply. The second is whether, in relation to retransmitted material, a broadcaster is a subordinate distributor.

There are three matters which clearly emerge from the decisions with respect to innocent dissemination. The first is that printers have been treated as not coming within the rule; the second is that the onus is on a distributor who claims the benefit of the rule to establish that he or she neither knew nor ought to have known of the defamatory character of the material in question; the third is that a distributor who establishes that he or she is an innocent disseminator is not liable because, for the purposes of the law of defamation, he or she did not publish the material in question.[82]

128 Her Honour then analysed the material this way:

For immediate purposes, all that need be said is that the evidence is that, unlike printers in earlier times, broadcasters who retransmit televised material do not necessarily acquire any knowledge of that material. Nor is it relevant for present purposes to consider whether a broadcaster ought to know what is being retransmitted. That is a question to be answered by reference to the circumstances of the case. Thus, it cannot be said that, as a matter of fact, the rule as to innocent dissemination cannot apply to those who retransmit televised material. Accordingly, it is necessary to consider whether a broadcaster is, in relation to the retransmission of televised material, a secondary distributor.

The respondents contend that the rationale underlying the concept of "secondary distributor" is that the person in question does not participate in "the production, selection or composition of the matter" and does not have "the ability to exercise control or supervision over the material which makes up the matter published". The difficulty with that submission is in relating it to some aspect of the law of defamation or, more precisely, some aspect of the law relating to publication, it being clear, as already indicated, that, for the purposes of the law of defamation, innocent dissemination does not constitute publication.

It has been suggested that innocent dissemination does not constitute publication because of the requirement in earlier times that publication be malicious. That may be so. However, the concept of innocent dissemination - a concept which was developed to take account of the practices and technology of the day with respect to the publication of printed matter and which now falls for consideration in the light of modern technology - necessarily directs attention to the present law relating to publication.[83]

In my view, it ought now be accepted that one who publishes by authorising a communication is not a subordinate distributor. Conversely, in my view, it ought also be accepted that one who does not authorise the communication but participates in it in some other way is a subordinate distributor and entitled to rely on the defence of innocent dissemination. To put the matter that way is simply to put a person who only participates in a mass communication on an equal footing with one who communicates defamatory matter to an individual. As already indicated, a person who communicates defamatory matter to another is liable only if the communication is intentional or negligent.

There can be no doubt that Channel 7 authorised the retransmission to its viewers by its servants or agents of the material which was defamatory of the appellant. Without its authority, the material would not have been retransmitted. And it is sufficient that it authorised the retransmission to its viewers of whatever was transmitted by Channel 9 without regard to its contents. Having authorised its retransmission, Channel 7 published the material in question. It cannot rely on the defence of innocent dissemination.[84]

129 Justice Gaudron dealt with the issue very briefly.[85]

130 Thompson is significant, first, because the plurality and Gaudron J, not for identical reasons, concluded that a re-transmitter of electronic material was to be regarded as a primary publisher of that material in certain circumstances. In that case, according to the plurality, the defendant had the ability to control and supervise the material which it broadcast. Having that ability, it chose to re-transmit the material immediately. In doing so, it was to be taken to have authorised the content of the material. Justice Gaudron approached the matter squarely upon the proposition that a person who publishes by authorising a publication is not a subordinate publisher. Thompson is secondly significant because it was not said in either of the judgments to which we have just referred that a re-transmitter of electronic material was incapable of being a secondary publisher, and hence able to rely upon innocent dissemination. It was accepted that the defence might run in the case of a distributor of electronic material if the Emmens test was satisfied. In the instant case, it was not. What the defendant had published, as a primary publisher, was evidently the entirety of the material which had been broadcast.

131 There has been academic criticism of the judgments of the plurality and of Gaudron J in Thompson.[86] But the short point, so far as the decision might be relevant, is that it binds us.

The Defamation Act 2005

132 We must say a little about the Defamation Act 2005, which applies to this proceeding.

133 First, the operation of the general law - that is, the common law and equity - is unaffected except to the extent that the Act either expressly or by necessary implication provides otherwise.

134 Second, 'matter' is defined by s 4 to include -

(b) a program, report, advertisement or other thing communicated by means of television, radio, the Internet or any other form of electronic communication.

135 Third, by s 8 -

A person has a single cause of action for defamation in relation to the publication of defamatory matter about the person even if more than one defamatory imputation about the person is carried by the matter.

136 Fourth, the Act is replete with references to 'publish', 'publication' and 'publisher'. But there is no general definition of any of these terms. There is a definition of 'the publisher' for the purposes of Division 1 of Part 3 of the Act, but that does not lead anywhere for present purposes. In the upshot, recourse must be had to the common law to determine whether a defendant has published defamatory matter respecting a plaintiff.

137 Fifth, Part 4, Division 2 of the Act sets out a number of statutory defences which are available to a defendant in a defamation proceeding. It is only necessary, for present purposes, to refer to s 32. Sub-section (1) of that section provides:

Defence of innocent dissemination

(1) It is a defence to the publication of defamatory matter if the defendant proves that -

(a) the defendant published the matter merely in the capacity, or as an employee or agent, of a subordinate distributor; and

(b) the defendant neither knew, nor ought reasonably to have known, that the matter was defamatory; and

(c) the defendant's lack of knowledge was not due to any negligence on the part of the defendant.

138 By sub-s (2), a person is a 'subordinate distributor' of defamatory material if it -

(a) was not the first or primary distributor of the matter; and

(b) was not the author or originator of the matter; and

(c) did not have any capacity to exercise editorial control over the content of the matter (or over the publication of the matter) before it was first published.

139 By sub-s (3) it is provided, without limiting the generality of sub-s (2)(a), that a person is not the first or primary distributor of matter merely because the person was involved in the publication of the matter in any of the eight circumstances thereafter set out. Thus:

(a) a bookseller, newsagent or news-vendor; or

(b) a librarian; or

(c) a wholesaler or retailer of the matter; or

(d) a provider of postal or similar services by means of which the matter is published; or

(e) a broadcaster of a live programme (whether on television, radio or otherwise) containing the matter in circumstances in which the broadcaster has no effective control over the person who makes the statements that comprise the matter; or

(f) a provider of services consisting of -

(i) the processing, copying, distributing or selling of any electronic medium in or on which the matter is recorded; or

(ii) the operation of, or the provision of any equipment, system or service, by means of which the matter is retrieved, copied, distributed or made available in electronic form; or

(g) an operator of, or a provider of access to, a communications system by means of which the matter is transmitted, or made available, by another person over whom the operator or provider has no effective control; or

(h) a person who, on the instructions or at the direction of another person, prints or produces, reprints or reproduces or distributes the matter for or on behalf of that other person.

140 It can be seen that most of the subparagraphs address matters previously dealt with by the common law. So, sub-paragraph (a) refers to the Emmens situation and sub-paragraph (b) to that which arose in Vizetelly. Sub-paragraph (d) picks up the situation of what may be called 'offline intermediaries', who convey a document containing defamatory matter from sender to recipient. Sub-paragraph (e) may have a connection with the analysis in Thompson, denying subordinate publisher status to an electronic broadcaster in some circumstances.

141 Next, sub-paragraphs (f) and (g) resemble, but are not identical with, s 1(3)(c) and (e) of the Defamation Act 1996 (UK).

142 Sub-paragraph (h) is a statutory recognition of the fact that the modern day printer, in the course of his or her employment, no longer manually typesets that which is to be printed.

143 Because s 32 treats innocent dissemination by a secondary publisher as a defence, it seems probable (although the matter was not argued in this Court) that the 'no publication' analysis is no longer available in a proceeding governed by the Act. The position may be different in the United Kingdom in the case of the persons, inter alia, to whom s 1(3)(c) and (e) of the Defamation Act 1996 apply. That section was referred to in some of the authorities respecting internet publication to which we will later refer.

144 One other matter should be mentioned. Section 32 has nothing to say about the type of situations which arose in Byrne and Urbanchich. They were not instances of secondary publication at common law; and they have not been converted by s 32 of the Act into something different. The principle developed in those cases remains an operative part of the common law.

145 Google contends that the plaintiff has no real prospect of succeeding because, inter alia, he has no real prospect of establishing that it published the allegedly defamatory matter. We have previously identified the impugned matter as being, in the case of the images matter, the composite of the search terms and the images compiled in response, noting that so to describe the images matter is the basis most favourable to the plaintiff. In the case of the web matter, we have described the allegedly defamatory material as the composite of the search terms and the results which the Google search engine produced, this again being the most favourable description of the impugned material from the plaintiff's standpoint.

146 The allegedly defamatory material was produced, as the plaintiff would have it, by accessing the Google.com.au, Google.com and Google.com.de websites, and typing in the various search terms.[87]

147 Immediately, a context is established: (1) the world wide web; (2) a particular 'search engine' website; (3) the ability of any internet user to access that website, using a web browser to input search terms; and (4), the form of the search engine's response to the terms propounded by the user. We think that there would scarcely be an internet user in Australia (or in the 189 countries where the Google search engine is used)[88] who would not recognise that context.

149 First, any user of a search engine will know of the extreme speed with which search results are produced. In the present case, for instance, page five of annexure A shows that about 78 600 results were produced in 0.35 seconds, for page six about 74 900 results in 0.32 seconds, for page two of annexure B about 378 000 results in 0.29 seconds, for page three about 1 490 000 results in 0.43 seconds, and for page four about 980 000 results in 0.19 seconds.

150 Second, it is inevitable that a user will understand - because of the speed with which search results are generated, and the number of search results produced - the enormous scale of the search which has been made; a search which could not possibly have been performed manually.

151 Third, and to take but two examples, an internet user who, using the Google search engine for the first time, went to the images section on a particular occasion, inputted the words 'melbourne criminal underworld photos', and received in response a compilation of images including some of known or reputed criminals but also images of the late Marlon Brando, a tram, actors in a film and a solicitor;[89] or, who received, in response to the same search words on a different occasion, a compilation including images of known or reputed criminals but also images of a former chief commissioner of police, a murder victim, a crime reporter and the Google logo,[90] would inevitably give thought to just what relationship there could possibly be between the words inputted and the compilation produced, and very probably perceive a disconnect between the images and the search terms; whilst a repeat user would inevitably, in our view, recognise - without necessarily understanding why it is so - that the search results in their entirety did not reflect the meaning of the inputted words considered as a phrase.[91]

152 The last three matters we have mentioned bear upon the issues of publication and capacity to defame. They invite some consideration of what the internet is, and how it operates; and also consideration of some features of the Google search engine. Consideration of each of those matters is also pertinent because a number of cases, worldwide, have considered the operation of the principles of defamation law with respect to aspects of the way in which the internet operates. But, as will be seen, those aspects differ to an extent one from the other; and all of them differ to some extent from the particular aspect of internet use upon which the plaintiff seeks to rely. We consider that, unless the differences are appreciated, the potential for jumping to the conclusion that the answer to the matters for our determination is to be found in authorities involving different aspects of internet use could be very real. So we will describe various aspects of the working of the internet at a basic level. Some of those aspects, indeed, emerge from descriptions in authorities to which we will refer. There is also the affidavit of Vaughn Beckett Madden-Woods affirmed 19 December 2014 ('the Madden-Woods affidavit') and the Herscovici affidavit to which we have earlier referred. Texts conveniently summarise the import of the material contained in the authorities and the affidavits.[92] There could be no serious argument against the proposition that everything which we describe in the next section of our reasons is either a matter of general knowledge in the community, a matter already recognised by an earlier decision or decisions, or else a matter made clear by uncontradicted evidence in the present matter.

153 In 1997, in Reno v American Civil Liberties Union,[93] Justice Stevens of the United States Supreme Court noted that the internet had experienced extraordinary growth. He said:

The number of "host" computers - those that store information and relay communications - increased from about 300 in 1981 to approximately 9,400,000 by the time of trial in 1996. Roughly 60% of these hosts are located in the United States. About 40 million people used the internet at time of trial, a number that is expected to mushroom to 200 million by 1999.

154 His Honour's prediction of continued extraordinary growth is reflected in an estimate that 47 per cent of the world's population, and 81 per cent of the developed world's population, are internet users.[94]

155 According to the responsible United Nations entity, the International Telecommunication Union, the estimated number of internet users worldwide will have more than doubled from about 1.4 billion in 2007 to over 3 billion in 2016. Further, as at mid-2015 it was estimated by the Australian Bureau of Statistics that about 86 per cent of Australian households had internet access, this being, comparatively, an extremely high proportion of the population.[95]

156 The rapidly expanding size of the internet is demonstrable in other ways. One indication is the increasing number of webpages, each with its own URL. According to Madden-Woods, as at 2008 the Google search engine had located, in a manner which we will later describe, one trillion unique URLs. By mid-2012, this had grown to over 30 trillion unique URLs. By mid-2013, over 60 trillion unique URLs had been found. The witness did not provide any updated figure as to the number of unique websites as at December 2014, when he affirmed his affidavit. But it can readily be supposed that, at that time, the 60 trillion unique URLs present as at August 2013 had markedly increased again.

157 Another indication, not only of internet expansion but also of use of the internet, is Madden-Woods's averment that, at the time of affirming his affidavit, over 100 billion searches per month - that is, over 3.3 billion per day[96] - were being made by users of the Google search engine, of which approximately 500 million each day were searches that had never been made before.

158 What we have thus far said reveals the extraordinary growth in the internet, and internet use, in the period of a little over 20 years that has elapsed since Justice Stevens wrote his opinion in Reno v American Civil Liberties Union; but it says nothing as to the different aspects of the internet.

159 The internet is a global network of computer networks which support communications services using the Internet Protocol.[97] Today, there are very many possible uses of the internet, and it is beyond question that a large section of the Australian population regularly engages with them. The NBN project was presumably conceived as a means of facilitating internet use still further. Email, Facebook,[98] Twitter, blogs, YouTube, iTunes, TripAdvisor, Google Maps, eBay, Gumtree, Amazon, Ticketmaster and Paypal are well-known names or descriptions of just some of the services and sources of information which are accessible via the internet. Internet banking is another.

160 Riordan describes an internet service provider ('ISP'), as the 'first class of network layer service'. It -

connects its subscribers to the internet by supplying telecommunications facilities and access equipment, such as modems and subscriber lines. When a subscriber requests or publishes content on a third party website, the packets pass through the ISP's network and are relayed to the remote host, which transports the response back to the subscriber.[99]

161 ISPs, he notes, regulate their relationship with subscribers by contract, often placing limits on the volume of data which may be downloaded, permitted uses of the connection, and the types of content which may be received and transmitted.

162 In some instances, an ISP may provide other services - for example, space on a server to publish a webpage.

163 How, then, does a user access a webpage? In the broadest terms, it is by the use of a so-called web browser. Instances are Google Chrome, Safari (provided by Apple), Mozilla Firefox and Internet Explorer (provided by Microsoft). Essentially, the browser permits communication between the user and an identified webpage.

164 The world wide web permits human interaction via so-called social platforms. Facebook, Twitter and Google+ are platforms of that kind. They publish user-created content, available in some instances to other users connected to the particular platform - 'friends' - and in other instances to the 'user world' at large. To call them 'platforms' is not to suggest that they are not accessible in the form of websites.

165 Pausing for a moment, it can be said that the Facebook and Twitter webpages (to take two well-known platforms) have about them something of the character of a noticeboard, owned or controlled by another, on which a person can put up a note, story, video or image. The circumstances differ from those considered in Byrne and Urbanchich, however, because the medium is different, and also because the quantity of material which is posted on those platforms each day is extraordinarily large. Another difference is that the platform host provides the platform, and often encourages its use. In neither Byrne nor Urbanchich was the place on which the material was affixed provided by the defendant in order that material might be affixed to it.

166 Other platforms permit account holders to upload material, and to post comments. YouTube and Instagram are examples.

167 The term 'publishing services' describes, inter alia, blogging services - the sites Blogger.com and Wordpress are examples - and discussion fora. Blogging services host publications authored by their users.

168 An individual or a company may also establish his, her or its own blog, whether standalone or part of a larger website.

169 Blogs, on which individuals post stories typically not limited by a maximum number of characters (by contrast with Twitter), generally permit responses from other internet users. It would be well nigh impossible to catalogue the matters which bloggers apparently consider will be of interest to others.

170 Gateways collate, index, and distribute hyperlinks to third parties' internet content. Search engines, portals, directories, and RSS[100] are the most common examples. Whilst these services employ various means to locate and rank relevant material, they are united by their reliance upon automated tools and algorithms to parse, store, and query large volumes of data authored by others.[101]

171 The Google search engine is a prime example of a gateway website.

172 Gateway sites are a practical necessity because the world wide web, as will be evident from our very basic description of it, comprises an almost incalculably large number of unique sites, each with its URL - with a secondary URL for each page on the particular site. Moreover, and importantly, the web is not static. That is so for two reasons. First, the number of webpages is constantly expanding. Second, the content of existing webpages changes - at least in very many instances. So, to take an example, a search of the webpage of a daily newspaper will reveal a constantly changing content. So it can be said that the overall content of the world wide web today will not be the same as the content of the web tomorrow; and, indeed, that its content this hour will not be the same as its content next hour.

173 It is evident that, unless an internet user knows the unique URL of the site which he or she wishes to access, the prospect of locating the site will be near hopeless. Hence the need - reflected by 3.3 billion searches each day using the Google search engine, and bearing in mind that there are other search engines, including Bing and Yahoo! - for a means of locating what the searcher wants.

174 Unlike a library, the web has no central control or cataloguing system. It is the absence of any overall cataloguing, the size of the world wide web, and the fact that a searcher can only access a particular page if its unique URL is known, that makes the use of a search engine greatly needed and extraordinarily common.

175 In Google Inc v Australian Competition and Consumer Commission,[102] a case concerned with Google's alleged breach of s 52 of the Trade Practices Act 1974 (Cth), the question was whether Google had engaged in misleading or deceptive conduct when search results consisted not only of 'organic' search results, but also sponsored advertisements. The particular issue was decided in favour of Google. For present purposes, however, we refer only to what was said by French CJ, Crennan and Kiefel JJ under the heading, 'The internet and the Google search engine':

1. The internet is a global network of networks of computers. Computers connected to the internet communicate with each other - requesting and receiving data - by means of a common language, the Internet Protocol. The World Wide Web is a vast system of interlinked documents ("web pages") which can be accessed by computers connected to the internet. Each web page has a unique address, or URL. An internet user who wishes to access a web page at a known address can access that web page by entering the address into the web browser on his or her computer.
2. An internet user who wishes to access a web page but does not know its address, or wishes to locate a selection of web pages relevant to a particular topic, is likely to use an internet search engine, like the Google search engine, in much the same way that a person who does not know the telephone number of a particular business, or wishes to contact a local provider of a particular product or service, might once have been likely to use a telephone directory.
3. The Google search engine allows internet users to search for web pages by entering search terms into a search field and clicking on a button marked "Google Search" ("the search button"). Google keeps and constantly updates an index of billions of web pages which enables it to respond to users' search requests. Google does not control the search terms entered by users of the Google search engine, or the material available on the internet.
4. During the period relevant to these proceedings, if a user of the Google search engine entered search terms into the search field and clicked on the search button, the Google search engine would display two types of search results: "organic search results" and "sponsored links".
5. Organic search results are links to web pages, which are ranked in order of relevance to the search terms entered by the user. The Google search engine always displays organic search results, and organic search results are always displayed free of charge. Google does not sell placement in its organic search results. Instead, the order of relevance of organic search results is determined by a complex proprietary algorithm developed by Google which is a function of many factors, including the content of each web page which Google has indexed, and the number and type of links between each of those web pages.[103]

176 We refer also to the following summary of the factual circumstances in the reasons of Heydon J:

1. This appeal concerns the business of the appellant, Google Inc ("Google"). That business depends on the World Wide Web ("the web"). The web is a vast system of linked documents to which access may be gained via the internet. Google may be described as a "website operator". Access to the web is obtained by using a "browser". That expression refers to the software used to navigate the web. A person wishing to make documents available on the web specifies an address known as a "Uniform Resource Locator" ("URL"). The URL appears in the address bar at the top of the screen which conveys the information contained on the web. The browser processes the URL to locate the information made available by a website operator. It then translates an underlying code into a webpage visible to an end user.
2. Google operates a free internet search engine. It is accessible on the internet at, among other places, www.google.com.au. Google's business reflects a relationship between Google, end users who conduct searches, and advertisers. Members of the public may conduct a search by entering a word-based query. In response, the search engine displays a results page. Two features of the results page are relevant to this appeal.
3. The first relevant feature of the results page is that it provides a list of links to webpages that may be of interest to the user. The list often runs into several pages. It may run into hundreds of pages. These links are called "organic search results". Each day many millions of search queries are conducted at www.google.com.au, each search taking a fraction of a second.[104]

177 Pausing for a moment, it can be seen that the interaction between a user and the Google search engine is quite different to the interaction between a user and Facebook or Twitter or the like. The user instigates the search by inputting search terms. The results which the search engine produces are a response which identifies webpages, or images from pages, or both. Those pages have been created by authors other than Google. Thus, Google neither compiles the search terms nor any webpages or images which are identified in response. This is different in kind to those websites which host text or photographs authored by users - that is, Facebook, Twitter or the like. In those situations, a question informed by the particular circumstances may arise whether it is technically possible for the platform provider to remove a defamatory post (despite the logistical problems of a huge number of postings each day), if informed by an aggrieved person of an allegedly defamatory posting.

178 As Heydon J said in ACCC, a search on the Google search engine involves entry of a word-based query; and as French CJ, Crennan and Kiefel JJ said in the same case, 'Google keeps and constantly updates an index of billions of webpages which enables it to respond to users' search requests'.[105] As their Honours also noted, search results are ranked in an order of relevance determined by a complex algorithm 'developed by Google which is a function of many factors including the content of each webpage which Google has indexed, and the number and type of links between each of those webpages'.[106]

179 Their Honours' succinct descriptions mirror passages in the Madden-Woods affidavit to which we have already referred. But we should expand upon these matters a little.

1. A search engine is an automated information retrieval system designed to allow a user to navigate the extensive information on the Web by the use of user-designed queries. A simple query might consist of two search terms, or keywords, such as "orange" and "apple". (In response to a user query, the Google Search Engine will use the logical connector "AND" by default between all the keywords that are part of that user query.)
2. More complex queries are possible. For example, a search query "orange apple", with the addition of quotation marks around the terms, instructs the Google Search Engine algorithms to strongly prefer results matching the exact phrase "orange apple" over results matching those same terms occurring further apart from each other in the webpage. A search query orange-apple.pear, with the addition of punctuation between the terms, similarly instructs the Google Search Engine algorithms to strongly prefer results matching the exact phrase "orange apple pear'", over results matching those same terms occurring further apart from one another in the webpage. That is to say, the Google Search Engine treats the search query orange-apple. pear equivalently to the phrase search query "orange apple pear", where the individual search terms are typed inside quotation marks.[107]

181 As to the way in which the Google search engine responds, Madden-Woods avers:[108]

1. Search engines, like the Google Search Engine, use computer algorithms to make predictions about what webpages among the trillions of pages constituting the Web are most likely to be of interest to the user. ...
2. Typically, a search engine returns results in the form of hyperlinks to webpages on the Web that the search engine has previously (a) located on the Web and (b) catalogued in an index. Typically as well, a search engine does not provide the user with just one result, but rather with a list of results potentially relevant to the user's query. The number of results returned by a search engine in response to a particular query might be in the order of 10 or 20, or it might be in the order of several thousands. How many results are returned depends on the nature of the query entered by the user and on the volume of available information on the Web. Given that the number of potentially relevant results may be in the order of several thousands, a search engine must use algorithms to rank those results by its prediction of the relevance of the results to the user's query.

182 Next, respecting Web Search and Image Search, Madden-Woods avers that:[109]

1. In very simple terms, the way in which Google Web Search and Google Image Search operate is very similar. The user enters some keywords as search terms. In response to the user's request for a search to be performed, the Google Search Engine takes those keywords and, by reference to a huge Index of the Web it has created, compiles a list of results. Those results are ranked, meaning they are ordered according to which ones the user will most likely be interested in, as determined by a computer algorithm. In that order, the browser on the user's computer will present those results to the user.

183 Specifically as to Image Search, Madden-Woods affirms:[110]

1. In the case of Google Image Search, only images are returned as results. The images that are returned as results are ones that are present on webpages that have been selected as relevant to the search terms that the user entered as part of the search query. By way of example, if a webpage about railway stations of the world contains text and images relating to both Flinders Street Station in Melbourne and Grand Central Station in New York, images of both stations will be identified as relevant in response to the user entering the query "Flinders Street Station" in Google Image Search. This is because those images are present on a webpage that is selected as relevant to that search query. With respect to the set of images in a webpage identified as relevant, Google Image Search will also rank those images according to their relative relevance to the query, as determined automatically by algorithms.

184 An image included in a result is only the beginning:

1. Each image included in the results returned by a search performed using Google Image Search is a thumbnail version of an image file on the Web. That image file will have a unique URL. For each image that is returned as a result, some HTML code is associated to it. That HTML code contains additional information (known as "interstitial" content) about that image result, which the browser will display when the user clicks on that image.
2. More precisely, if the user clicks on the thumbnail version of the image that is included in the results returned by the search that has been performed, the user will be presented with a "black frame" with the following components (this is the "interstitial" content):

(a) the image selected (this is a larger version of the image than the thumbnail included in the results, and it is served from the website that stores that image);

(b) a title for the image, to which there will be associated some HTML code including the unique URL to the webpage where the image is located;

(c) a word or phrase which the algorithm has determined may assist the user to identify whether the image is relevant to the user's search query;

(d) the dimension of the image in pixels;

(e) a snippet from the webpage where the image is located;

(f) two buttons, "Visit page" and "View image"; and

(g) thumbnails of other similar images.

Clicking on the large image, the "Visit page" link, or the title, will take the user to the webpage where that image is located. Clicking on the link "View image" will take the user to the webpage where the Google Search Engine located that image, and the user's browser will display a larger version of the image. The purpose of the interstitial content is to give more information to the user with respect to the image and its provenance.[111]

185 Pausing, it will be understood that, in the case of the images matter now complained of, the searcher took no step beyond obtaining the images results.

186 The explanation of the apparently incongruous images depicted on webpages relied upon by the plaintiff - a matter to which we referred at [151] above - is given by Madden-Woods as follows:

1. For example, page 12 of Annexure A of the Amended Statement of Claim (see exhibit "VBMW-15") shows, in the 3rd row, what I understand to be photographs of the Plaintiff (second-last photograph), and of Mr John Silvester (last photograph). When I performed a search on Google Image Search with the terms "john silvester",·"melbourne", "criminal" and "underworld", I obtained the results shown in exhibit "VBMW-9". When I clicked on the 2nd image in the 2nd row (this image seems to me to be identical to the .image of Mr John Silvester as shown on page 12 of Annexure A), I obtained the interstitial page shown in exhibit "VBMW-10'". When I clicked on "Visit page", the browser on my computer displayed the webpage with the URL http://www.australianspeaker.com/speaker1386-John-Silvester, a copy of which is exhibit "VBMW-11". That webpage contains the search terms "john silvester", "melbourne", "criminal" and "underworld", as highlighted in the exhibit.
2. When I performed a search on Google Image Search with the terms "trkulja", "melbourne", "criminal" and "underworld", I obtained the results shown in exhibit "VBMW-12". When I clicked on each of the 1st, 2nd and 3rd images in the 1st row (these images seem to me to be identical to the image of the Plaintiff as shown in the 3rd row on page 12 of Annexure A), I obtained the interstitial pages shown in exhibit "VBMW-13". When I clicked on "Visit page" for each of the three interstitial pages, the browser on my computer displayed the webpages with the URLs http://www.afr.com/p/national/melbourne\_man\_wins\_in\_google\_case\_vAGdjwqAJNsOt2ZXsOhGHL, http://www.smh.com.au/business/the-biggest-evil-milorad-trkula-wants-to-be-removed-from-google-20131205-2yrqj.html and http://www.telegraph.co.uk/technology/google/9648750/Google-loses-David-and-Goliath-defamation-case.html, copies of which are exhibit "VBMW-14". Those webpages contain the search terms "trkulja", "melbourne", "criminal" (or "crime") and "underworld", as highlighted in the exhibit.
3. By reason of the changing nature of the Web, it is impossible for me to say with certainty that the webpages from which the images of the Plaintiff and of Mr John Silvester (shown in the 3rd row of page 12 of Annexure A) originated are the same webpages as the ones shown at exhibits "VBMW-11" and "VBMW-14". The three searches were done at different times - the one that led to the printout at page 12 of Annexure A was done on 8 August 2013, the other two searches were done on 30 May 2014.
4. However, what the searches done on 30 May 2014 (see "VBMW-9" to "VBMW-11") do show is that the image of an individual (for example, Mr John Silvester) can be included in the search results for queries that include the search terms "criminal", "melbourne" and "underworld" simply because those terms appear in the webpage which contains that image. There could be nothing at all in the content of that webpage which imputes criminality or being part of the "underworld" of Melbourne.[112]

We append as Annexure 3 to these reasons copies of exhibits VBMW 9, 10, 11, 12, 13 and 14 to the Madden-Woods affidavit in order that the above cited paragraphs may be readily understood.

187 The witness next deals with a web search, of the kind which produced pages one to five of Annexure B:[113]

1. In the case of Google Web Search, the main type of result that is returned generally will consist of the title of the webpage, the URL and the snippet for that webpage. By way of example, page 2 of Annexure B of the Amended Statement of Claim (see exhibit "VBMW-16") has, as the second result, the following parts:

(a) Title - in the example, the title of the webpage is "Trkulja v Yahoo! - Defamation Watch". A title is also a hyperlink, which the user can click to reach that webpage.

(b) URL - In the example, the URL is shown, in part, as "defamationwatch.com.au/music-promoter-wins-225000-trkulja-v-ya ...".

(c) Snippet - a snippet is a small amount of text, derived from the textual content of the webpage, which is shown to the user as an aid for assessing the relevance of the webpage. If the search terms appear in the snippet, they are represented in bold. In the example, the snippet reads:

22 March 2012 - 7" search service on a website entitled "Melbourne Crime" .... To the right of the article was a large photo of Trkulja and then an article: ... the plaintiff is such a significant figure in the Melbourne criminal underworld that events ...

1. Also on this document, the first result consists of 15 images that have, as the heading, the text "Images for melbourne criminal underworld photos" (followed by the text "Report images"). The images that are shown have been evaluated to be, by automated computer algorithms, the most relevant among all the images contained in the webpages that are the most relevant results. The text "Images for [keyword terms]" is a hyperlink. When the user clicks on this hyperlink, the user is instructing Google Image Search to perform a search for images using the same keyword terms that the user had entered in Google Web Search.

We append as Annexure 4 to these reasons a copy of exhibit VBMW 16, so that the above paragraphs may be readily understood.

188 There was reference at [20] in the judgment of French CJ, Crennan and Kiefel JJ in ACCC to Google maintaining and constantly updating an index of billions of webpages which enables it to respond to users' search requests. In his affidavit, Madden-Woods explains indexing, and also the anterior and subsequent steps of 'crawling the web' and 'ranking and returning the results to the user' being performed in an automated way in accordance with complex algorithms designed or written by Google employees.

189 The overall object of the programs is to permit information which may be useful to the searcher - bearing in mind that there are around 3.3 billion searches per day worldwide - to be identified extremely quickly from the content of the almost incalculably large number of URLs.

190 Crawling is undertaken in accordance with a computer program known as a 'web crawler' or 'robot'. Its purpose, in short, is to repetitively crawl the web, accessing and receiving information contained on the trillions of webpages identified by URLs. The information will commonly contain text,[114] and there may be image or video files.

191 Madden-Woods describes the way in which the web crawler deals with an image this way:

The Web crawler does not read (parse) a media file, such as an image. Instead, a copy of the image is retrieved by the Web crawler and a copy is stored, separately from the other data for that webpage. It is later processed to identify information about that image such as the dimensions, measured in pixels, and other visual attributes such as dominant colour, or whether the file is likely to be a photographic image or a drawing.[115]

192 Web crawling is a constant process. Madden-Woods describes it this way:[116]

1. The crawler program is constantly visiting and processing webpages on the Web. The process is performed with the use of a large number of computers that run the Web crawler algorithms to visit a large number of web pages at once. The Web crawlers for the Google Search Engine visit more than 20 billion webpages in a day. A large number of computers also constantly perform the algorithms that analyse the images.
2. The crawler program, a fully automated computer program with no human intervention, determines which websites to crawl, how often to crawl the websites and what information is collected about those websites. For example, the frequency with which a particular webpage is crawled, as determined by the crawler program, depends on the relative importance of the webpage determined by an assessment of factors such as how frequently the web page is updated, how popular the website where that webpage is located is, and how many requests can reasonably be handled by that website's host computer in a given timeframe. Important pages may be crawled for new data every few minutes to few hours. Less important pages may be crawled at intervals of weeks or months.
3. Every time a webpage is re-crawled and new data detected, the stored data relating to that webpage is updated. If, on a re-crawl, the Web crawler detects that a webpage is no longer available, that information is passed on to the indexer program, so that indexing to that webpage will no longer occur and, as a consequence, that webpage will no longer be returned as a result in any search using the Google Search Engine.

193 We pause to observe that the reference at [97] in Madden-Woods's affidavit to there being no human intervention is true of the process itself. The point made by Beach J in Trkulja (No 5), and by the judge below, is that the process which leads to results being provided, though automated, is the consequence of computer programs designed and written by Google employees.

194 We go to indexing, which is done by another computer program. The purpose of indexing is to transform the data from the stored webpage 'into a form that is more easily searched by computer algorithms'.[117] Madden-Woods describes the operation of the program this way:

1. The indexer program builds a list of every webpage that contains a particular word, say "orange". It does this for every word found during the crawling stage. Also, each webpage that is indexed is given a unique ID that can be looked up to find the unique URL for that webpage. As each webpage is indexed and parsed for all the words it contains (say, the words "orange", "apple", and so on), the unique ID for that webpage is added to the list of unique IDs that is created for each word contained in that webpage.
2. The words on the webpage being indexed are combined with the information from all other webpages that have been indexed. The index is refreshed constantly as the Web crawler crawls the Web.
3. The resulting index contains each word and a list of the unique IDs that relate to the webpages that contain that word. Separately, the correspondence between each unique ID and the unique URL for the webpage is also stored. In the resulting index, the word "orange" will appear next to a long list of unique IDs, one for each webpage on which that word appears. Looking up the index by the word "orange" will return the long list of unique IDs. Looking up each unique ID in that list will return the unique URL for that webpage.
4. The indexer program also notes many different aspects of a web page, such as the date it was published, whether the page is comprised of text, images or video, and other aspects, such as whether the webpage appears to be a news article. This information will be used later in the ranking stage, in ways that are proprietary and confidential. By way of example, however, I can say that if the webpage is a news article, it may be considered more highly relevant for some user search queries.[118]

195 Madden-Woods refers to a particular aspect of the indexing program referable to images:

1. In respect of an image that is part of a webpage, an algorithm that is executed separately to the indexer program attempts to pair text on that webpage with that image. An example of text likely to be highly relevant to the image is a caption, which can be specified for that image using HTML code. Another example is the file name for the image. The metadata for an image, for example, the metadata for a photograph, is also likely to be highly relevant. Other clues are used by that algorithm to try to pair relevant text on the webpage with the image, such as if the text is, in the HTML code, in close proximity to the URL for the image.
2. The next thing that happens at the indexer stage is that that paired text is indexed for that image in that webpage. The paired text is indexed in the same way as any other text found on a webpage.
3. The information about the image, such as dimensions and visual attributes, that was the result of processing by the crawler program is also entered into the index for that image.[119]

196 The next step is production of search results, initiated by a user's inquiry. Again, results are produced by the operation of algorithms designed and written by Google employees. Thus:

1. When a user enters a query in either Google Web Search or Google Image Search, the words from that query are evaluated, by a series of algorithms, against the information in the index created by the indexer program, in the form in which the index is at that precise point in time. The ultimate result, in the case of Google Web Search, is a list of links to webpages, presented to the user as search results ranked according to relevance, as estimated by the ranking algorithms. In the case of Google Image Search, the ultimate result is a list of images ranked according to relevance.
2. More specifically, the ranking program accesses the index created by the indexer program for each of the search terms entered as part of the query, and it looks up the unique IDs for all web pages that contain the term or terms entered by the user. The program then converts the IDs into URLs that relate to the relevant webpages on which the word was found. For common words, this will be a very long list of URLs.
3. For the list to be useful to users, the ranking program then uses over 200 unique "signals", or clues, to attempt to identify what results the user is most likely looking for. Some signals are more important than others and so have a larger impact on the score. Essentially, the ranking algorithm attempts to give higher scores to webpages that are more relevant to the search terms that the user chose. The results, links to webpages, will then be presented to the user according to their relevance as predicted by the Google Search Engine, with the most relevant result presented first.
4. The exact algorithm is proprietary information and highly confidential as it is a major differentiator between the Google Search Engine and the search engines provided by competitors. Some examples of signals which I can publicly identify are:

(a) the number of times one or more of the user's search terms appear on the webpage, as indexed by the indexer program;

(b) how often other webpages link to that webpage, and the importance of the linking webpages (this signal is known as PageRank);

(c) how recently the content of that webpage was published or updated (freshness);

(d) evaluating the order in which the search terms appear on a webpage.

...

1. For typical search terms, there may be thousands or millions of webpages with relevant information. For the search tool to be of utility to the user, the ranking algorithm attempts to predict which results will be most relevant to the user's query and provides them higher up the list. Because the ranking is performed so quickly and over so many pages, it cannot be done by a human using a subjective decision-making process. The algorithm ranks the pages by allocating, for each URL that appears in the results, a score. The higher the score, the higher that URL will appear in the results.
2. By way of explanation of a specific "signal", webpages most relevant to the search "Paris Hilton" are likely to be different from webpages most relevant to the search "Hilton Paris", even though there is only one list of web pages containing both of these terms that will be returned from the index. Google Web Search and Google Image Search make use of the order of the search terms in ranking the results, and so the search results for these two searches will differ.
3. Other "signals", used as part of the ranking algorithm, are more closely related to the particular user making the search query. One is the use of country-specific domains from where the Google Search Engine can be accessed by a user. (discussed below). Others are even more personal to the user, such as the user's previous search history (discussed below).
4. These personalisation "signals" mean that when two people conduct the same search (that is, they enter a search query using the same search terms in the same order) they may get different search results or the search results may appear in different orders.[120]

197 With respect to the results of an image search, Madden-Woods avers the following:

As with webpages, there are many different "signals" that are applied in the ranking stage of images. This is proprietary information and highly confidential. Some examples of signals which I can publicly identify are:

(a) a large image might be more relevant than a small image;

(b) an image located closer on the webpage to the search terms may be more relevant than an image that is located further away;

(c) a photographic image might be more relevant than a drawing if the query included the term "photos".[121]

198 Other aspects of the ranking system are referred to by Madden-Woods. Thus:

1. The Google Search Engine is customised for a number of countries and regions across the world. Country specific sites are a convenient way for Google Inc. to provide relevant and country specific information for users living in different countries. The Google Search Engine accessed by a user from google.com.au will provide search results that are most relevant to that user if he/she is in Australia. By way of example, if a user searches the word "mcdonalds" on google.com.au, the first result will be for mcdonalds.com.au, which is the Australian website for McDonalds. If a user were to search the word "mcdonalds" on google.com, the first result will be for mcdonalds.com, which is the US website for McDonalds. Similarly, a search on google.co.uk will return mcdonalds.co.uk as the first result.

1. The Google Search Engine is, in some cases, able to use the search history of the user to rank more highly certain webpages that, on the basis of that search history, are evaluated as more relevant to that user.
2. By way of example, suppose a user has done searches in the past for "Australian news websites", and that, in respect of those searches, the user always chose the URL for the Herald Sun website among the results being returned, even though it appeared as, say, the 6th or 7th result in the list. The Herald Sun website will, in subsequent searches by that user, be ranked more highly by the Google Search Engine, and may appear as the 1st result in the list.
3. There are two ways in which the Google Search Engine is able to access and utilise the search history of the user. First, when the user is signed into the Google system and the user has the feature "Web History" enabled. Secondly, by means of an anonymous cookie stored on the user's computer and linked to the browser, being a cookie which allows up to the prior 180 days of searches associated with that anonymous cookie to be maintained as search history. Personalisation by reference to a user's search history can affect the Images returned as results when that user uses Google Image Search.[122]

199 Madden-Woods also affirms, consistently with those parts of his affidavit to which we have already referred, and consistently also with Heydon J's reference in ACCC to the search being a 'word based query', that:

While the Google Search Engine has the information that certain words (for example, "orange" and "apple") appear somewhere in certain web pages, as part of the text of those web pages, it treats those words as strings of letters, not as words with assigned meanings. The Google Search Engine does not know what a particular webpage containing the words "orange" and "apple" might be about. Similarly, while the Google Search Engine has the information that, for example, an image identified by a unique URL is a part of a given webpage, and while the Google Search Engine attempts to link certain words in that webpage to the image, it treats those words as strings of letters and it does not know what the image might be about.[123]

200 Because the giving of notice by the plaintiff (putting to one side for the present the form of the notice) was relied upon both by the judge below and in argument in this Court by the plaintiff, we will refer to what Madden-Woods affirms in that connection. Thus:

1. Save for very limited exceptions, Google Inc. does not alter by human intervention the search results that are returned, in a fully automated manner, by the Google Search Engine.
2. The exceptions to the rule that Google Inc. does not alter by human intervention the search results are as follows:

(a) what are known as "legal removals" (discussed below);

(b) not returning a URL as part of the search results temporarily, where the web page corresponding to it publishes information which if discovered would imminently endanger a person's life, and that fact is brought to the attention of Google Inc.;

(c) not returning a URL as part of the search results, where the webpage corresponding to it is determined by Google Inc. to be "webspam", meaning that the webpage fails to meet Google Inc.'s established and published web master guidelines (an example of "webspam" is a webpage that presents useful content to the Web crawler, but which presents different - perhaps shocking or malicious - content to a user who accesses the webpage); and

(d) in the very rare circumstance of issues relating to security or stability, such as when the user's search query or a particular URL causes the Google Search Engine to crash.

...

1. "Removals" is a term used by Google Inc. for the processing of requests from users, courts, government or law enforcement, for Google Inc. to remove a webpage from the results that may be returned in relation to search queries. If a webpage is removed as a possible result, the unique URL for that webpage will no longer appear in the results in respect of any search query, made by a user from a specific domain for the Google Search Engine, that would otherwise have included it.

1. Although Google Inc. has no control over the Web, the Web crawler may visit, and the indexer program may index, a webpage that contains material that a person claims infringes their rights, such as copyright. As discussed above, an indexed webpage will appear as part of the search results if the ranking algorithms assess the content of that webpage to be relevant to the query that the user has entered into the Google Search Engine.
2. The next point is that the Google Search Engine is not capable of evaluating the content of a webpage, by which I mean the meanings that are conveyed by the words and images used on that webpage. Nor is the Google Search Engine capable of making a judgment call as to the topic or topics addressed by the content of that webpage. The Google Search Engine is unable to discern the standpoint of the author of a webpage on a particular topic. The Google Search Engine is unable to evaluate whether statements made in a webpage are true or not. Not only is the Google Search Engine not capable of doing the above, in my opinion, given the current state of the art in computer engineering, word, voice and image recognition, it is not feasible to design any automated system, on which a search engine might be based, which could determine those matters. It is well beyond the current state of the art to contemplate an automated system that could determine whether statements made on a webpage are true or not.
3. The Google Search Engine cannot discern whether the use of an image on a particular webpage is authorised by the copyright owner of that image. By way of example, one website owner may have obtained permission from the copyright owner of a particular photograph to use that photograph on a webpage, whilst another website owner might use the exact same photograph without having obtained permission from the copyright owner. The Google Search Engine has no ability to determine which use of the photograph, if any, is authorised.
4. It is also important to note that when a removal request is given effect by Google Inc., the entire website or domain is not removed from the search results, just the specific page complained of, corresponding to the URL communicated to Google Inc.
5. In terms of preventing a certain webpage from being returned as part of search results when a user enters a query in the Google Search Engine, this can be done only if Google Inc. is notified of the unique URL for that page. Importantly, even when Google Inc. effects a removal, this does not remove that webpage from the Web. As I said above, Google Inc. does not control the Web. Any user who knows the URL for that webpage can put that address directly in his/her browser and view the page. Similarly, existing hyperlinks in other webpages might take a user to that webpage.
6. If a request for a removal on legal grounds is received by Google Inc, it is considered by a legal team within Google Inc. If the decision is made to remove a webpage from the search results, the URL for that web page is given to a specific engineering team that will put into effect the removal request.
7. As part of the removals engineering team, I would receive from the legal team within Google Inc. a request to remove a webpage from search results returned for a particular country code domain of the Google Search Engine. For example, the request would be to remove the webpage from results returned from the Google Search Engine at google.com.au. I would then manually add the unique URL for that webpage to a list. During the ranking stage, the algorithm removes all the URLs found in the removals list from the list of possible results, before a ranked list of results is returned to the user's browser for display by the browser on that user's computer (or tablet or smartphone).
8. In my experience, it is only possible to remove a webpage from search results if Google Inc. is provided with the exact URL for the webpage containing the content complained of.[124]

201 In this Court, senior counsel for the plaintiff criticised, and sought to rely upon, what he submitted were inadequacies in Madden-Woods's affidavit respecting removals. That was one reason which he advanced why Google's application must fail. It is a submission to which we will return. We note, however, that he did not submit that the Court was not entitled to have regard to the affidavit.

202 The autocomplete feature of the Google search engine, in point because of pages five to seven of Annexure B, is addressed by the Herscovici affidavit. That witness affirms as follows:

1. The Autocomplete utility makes the Google Web Search and the Google Image Search functionalities more useful for the user. It does so by showing to the user possible ways of completing the query that the user is typing into the search box. As the user types first one character, then a second character, then a third character and so on (so much of the query as has been typed by the user at a given time is known as the "prefix"), the Autocomplete utility responds in real-time by showing to that user possible completions for each prefix.

1. The outcomes of Autocomplete depend on what the user does (in terms of the prefix typed into the search box), and on what that user and all other users of the Google Search Engine have done in the past (in terms of completed search queries).
2. The Web is continuously changing and growing. Each day, approximately 3.3 billion searches are performed by users of the Google Search Engine, and of those approximately 15% (or 500 million) are new searches, that is searches that had not previously been made by users of the Google Search Engine. Because users of the Google Search Engine enter different search queries at different times, depending, in part, on what is topical and hence likely to have resulted in the recent addition to the Web of relevant webpages, the universe of possible outcomes of the Autocomplete utility will necessarily vary with time. This means that in respect of a particular prefix entered by a user at a given time, Autocomplete may have shown certain predictions. If the same prefix is entered at some later point in time, Autocomplete may well show a very different set of predictions.
3. Rather than typing the search query in full, the user who has typed a given prefix can simply select one of the predictions being shown in the browser. The effect of this is that the user is directing the Google Search Engine to do a search (either Web Search or Image Search) using that selected prediction as the search query.

203 The statement by the judge below that 'provided that Google had an intention to publish the results that its search engine produced and of which Mr Trkulja complains, Google may still be found to be a publisher',[125] was founded on a conclusion expressed by Beach J in Trkulja (No 5).[126] Google's counsel submitted that his Honour had elaborated that basic proposition at various points in his reasons, culminating in his statement that

even if Google does not intend its search engine to produce results linking Mr Trkulja's image with those of convicted criminals, there is no suggestion that this outcome was anything other than the product of the intended operation of the search engine. Applying the reasoning of the High Court in Thompson, Google's ignorance of the specific material generated in response to any given user inquiry does not prevent it from being found liable as a publisher.[127]

204 Counsel submitted that the Thompson analogy was the only basis upon which the judge found that it was open to conclude that Google was a primary publisher - that being the plaintiff's case, and being what the judge must be understood to have decided was reasonably arguable.

205 Counsel further submitted that the judge had not addressed a post-notice argument - that is, founded on an extrapolation of 'primary publisher' cases of the Byrne and Urbanchich kind. We pause to note that, if the plaintiff's case was that Google was liable as a primary publisher, then secondary publication was irrelevant; so the Emmens and Vizetelly line of authority would be beside the point.

206 Counsel submitted also that a post-notice argument founded on the Byrne and Urbanchich line of authority could not sensibly have been maintained. That was because the web is constantly changing, so that today's response and tomorrow's response to a search term may be completely different. A static notice or poster was not comparable. What the plaintiff was really seeking was a mandatory permanent injunction to remove the possibility of a search return providing any link to any page on the web that might produce a particular image or combination of images. Such an injunction would not be granted.

207 Having made the point that the judge had founded his ruling on intention to publish the allegedly defamatory matter without consideration of a post-notification argument, counsel reiterated his submission that post-notification learning had nothing to say about search engine results. That was because -

What is necessary on all of the post-notification cases is a set of circumstances in which you can impute to the defendant an approbation, an approval, a ratification of the continuing publication of the particular defamatory terms, and that is impossible in this area.

208 Counsel explained the alleged error made by the judge with respect to intention to publish this way:

The intention that [the judge] founded himself on, which was that Google intends its search engines to publish material on the internet in response to a user inquiry, whatever that material might be, ... is a proposition pitched at a level that is too amorphous, too detached from reality and, on the authorities, incapable of establishing an intention to publish the defamatory material.

209 He submitted that only the judge below and Beach J in Trkulja (No 5) had held that an intention so expressed was sufficient with respect to publication of defamatory matter.

210 Counsel submitted that the foundation in Australia respecting intention is what was said by Isaacs J in Webb.[128] But whilst the judge below, and Beach J in Trkulja (No 5), had accepted the need to establish intention in the Webb sense, the judge below had erred in applying the 'intention' element by concluding that -

Provided that Google had an intention to publish the results that its search engine produced and of which Mr Trkulja complains, Google may still be found to be a publisher.[129]

211 Counsel cited and relied upon passages in the judgments in ACCC in support of his argument. He submitted that Google could not be regarded as having approved or affirmed or in some way taken responsibility for what appeared in response to inputted search terms.

212 Counsel also referred to and relied upon - (1) the decision of Eady J in Metropolitan Schools; (2) the judgment of McCallum J in Bleyer; (3) the decision of the Supreme Court of Canada in Crookes v Newton;[130] (4) the conclusion of Blue J in Duffy that Google was not the primary publisher of search results;[131] (5) the Court of Appeal decision in Tamiz that Google was not a primary publisher of posts on a blogging platform; (6) the leading judgment of Ribeiro PJ of the Hong Kong Final Court of Appeal in Fevaworks;[132] and the requirement stated by the New Zealand Court of Appeal in Murray v Wishart[133] that the defendant intend to publish the actual words used. Later, we will consider these authorities, and express our conclusions as to the propositions for which they relevantly stand.

213 Counsel contended that the decision in Thompson was not inconsistent with those authorities. The critical matter there was the control and ability to control the broadcast by the defendant, which was held to be a primary publisher. In the present situation, by contrast, it could not be said that Google would know what was going to be thrown up, in a fraction of a second, from the huge number of crawled and indexed webpages in response to an inputted search term.

214 Counsel submitted that the evidence showed that images and snippets were properly to be regarded as mere references indicating that there was some connection between the search terms and the contents of identified webpages on which the images appeared, or from which the snippets were derived. Google had no knowledge, or awareness, or means of control as to what was going to be returned.

215 Counsel made the point that, according to the evidence, it could not be said that a displayed image is relevant to the search term. It can only be said that the image was on a page which contained something relevant to the search term. (Thus, an image of the Chief Commissioner of police might be returned on an image search of, say 'Melbourne underworld criminals' because the image appeared on a webpage devoted to the Commissioner's zealous work in tracking down Melbourne underworld criminals). Counsel made the point also that when Madden-Woods made a search and clicked on a thumbnail image of the plaintiff, what was revealed was an account of the plaintiff's success in his earlier proceeding against Google.

216 As to intention, counsel submitted in summary that the plaintiff had no prospect of establishing that Google had the requisite intention to publish defamatory matter. It had 'merely provided a tool by which users can obtain references to places on the web, or items on the web that are connected in some fashion with the search terms entered by the individual'. This is all that a return designates.

217 Counsel met a suggestion from the Bench that Google benefited from setting up and making the search engine available to users - that is, by paid advertisements - by noting that -

An ISP service provider intends to profit by facilitating communications, and they intend to facilitate the communication of whatever users want to communicate. But that doesn't mean they have an intention to publish in the sense used in defamation law.

218 As to the autocomplete predictions upon which the plaintiff relies, counsel emphasised the importance of previous searches by, particularly, the particular searcher and by others using the particular combinations of words. At the heart of the submission was the unstated proposition that a searcher has an ability to manufacture - whether intentionally or otherwise - an autocomplete prediction which he or she then asserts is defamatory.

219 Senior Counsel for the plaintiff submitted that it would be wrong to shut out the plaintiff from pursuing his claim at this time. It would be premature to do so. The evidence upon which Google relied had not been tested before the judge. Moreover, it was relevantly a statement of what Google does, rather than what Google could do to prevent defamatory matter appearing in response to inputted search terms.

220 Counsel next submitted that there was no authority for the proposition that there is no publication of defamatory matter without intention. The issue was one receiving worldwide attention. Its resolution could only be properly addressed at trial, after completion of interlocutory processes, including discovery.

221 Counsel argued that the amended statement of claim does plead a post-notification case. He further submitted that a post-notification case or argument was in any event available to his client, and submitted that it would be a big step to set aside service when such a case was potentially available.

222 Counsel could not point to any part of the reasons of the judge below which analysed the issue in a post-notification context.

223 Further as to post-notification, counsel submitted that there were a number of cases which say that, after notification, Google may have been obliged to do its own continuous searching to prevent the risk of the return of further defamatory material in response to searches. He referred to European authorities which, he submitted, have required that Google comply with the law, being disentitled to demand a precise URL before responding to a complaint.

224 Further still as to the post-notification situation, counsel submitted that - (1) Eady J did not consider that issue in Metropolitan Schools; (2) McCallum J reserved the point in Bleyer; and (3) Duffy established liability with respect to post-notification publication. He referred also to the conclusion of the trial judge in Fevaworks that, after notification, the provider needed to take all reasonable steps to remove defamatory material.

225 We pause to note immediately a matter of considerable importance to the resolution of this application. The submission for the plaintiff that this was a 'post-notification case' focussed attention upon - (1) the ASoC; (2) the question whether the plaintiff's case was that the defendant was (a) a primary publisher, or (b) a secondary publisher; and (3), the judge's reasons. There was considerable debate in this Court as to - (1) whether the ASoC pleaded a post-notification case at all; (2) whether, and if so the way in which, notice had been referred to below; and (3), the basis upon which the plaintiff advanced his claim. The end result of the debate was clear: the plaintiff pursued his claim on the basis that the defendant was the primary publisher of the impugned matter. To the extent that notice might bear on such a case, the plaintiff was at liberty to rely upon it. The plaintiff did not advance a case that the defendant was a secondary publisher, and so notice was not called in aid in that connection.[134] That is so although authorities to which counsel referred in support of post-notification liability involved secondary publication.

226 As to what he submitted were the deficiencies in the affidavit material relied upon by Google, counsel submitted that:

[W]e say that the affidavit does not address, for example, the blocking of all references to a particularised name - and that is Madden-Woods' affidavit, paragraph 127 - or the blocking of all particularised snippets, paragraph 128; the capacity of Google to maintain a continuous search on a name; to identify future offending, that is 132, or not addressed in 132 or 162.

Mr Herscovici does not address the creation of indices and predictive text, autocompletes and the capacity to edit same following notification. That is Madden-Woods at 156, and Herscovici in 30, as I have taken the Court to. So all of these matters, we say, put that affidavit material into context and, we would suggest, prevent our learned friend from addressing the Court on the basis that Google has no control over these matters. That is a matter still to be determined.[135]

227 In the context of the web, and defamation, proceedings have been brought against the operators of -

228 Relatively infrequently, however, have proceedings been brought against the authors of webpages, or the makers of comments posted in discussion forums and on bulletin boards, blogs, or Facebook users. That is so even when the identity of the author is apparent or is perhaps ascertainable. Presumably, it is an entity of the kind commonly sued which is thought to be worth powder and shot. The fact that proceedings are commonly brought against such entities has thrown up distinct problems in defamation law.

229 Google[136] has been sued as a search engine operator, and as the operator of blog, social media and YouTube sites.

230 There have been a variety of circumstances in which defamation proceedings arising in connection with the internet have required consideration by the courts. Thus -

(1) applications by defendants for a permanent stay of the proceeding, or for summary judgment, or to set aside service outside the jurisdiction;

(2) application by a defendant for judgment non obstante veredicto;

From one situation to the other, the nature of the burden cast upon the moving party has varied.

231 The combination of the matters referred to in [227] and [230] means that close consideration must be given to what may legitimately be drawn from any part of a particular decision.

232 In the case of the Google search engine, the material alleged to constitute defamatory matter has been, in the particular case, one or more of -

(1) search results in the form of a URL plus hyperlink;

(2) the content of the page or pages accessed by clicking on a hyperlink; and

(3) rarely, autocomplete predictions or related searches suggestions.

233 Until the present case, so far as the researches of counsel and the Court reveal the situation, it appears that in the common law world image results simpliciter have not been relied upon as constituting defamatory matter. Further, until this proceeding, an autocomplete prediction has been relied upon as constituting defamatory matter only in Yeung and Duffy.

234 Counsel's industry meant that we were provided with decisions of courts of very different constitution - decisions of the Court of Final Appeal of Hong Kong and the Supreme Court of Canada, of the English and New Zealand Courts of Appeal, at Queen's Bench in England and Northern Ireland, of single judges in Victoria, New South Wales and South Australia, of an Associate Judge in New Zealand and a Deputy Judge in Hong Kong. None of those decisions bind us, but they have all addressed, in varying contexts, particular facets of internet defamation problems; and, allowing the significance of the varying contexts, have been helpful in defining and assisting resolution of the issues for our consideration.[137]

235 Plaintiff's counsel also provided us with decisions of European courts, not all of them relating to defamation proceedings. Those decisions appear to be overlaid, to some extent, by provisions and directives peculiar to European Union Member States. The position was complicated further by the fact that the decisions as provided to the Court were scarcely intelligible - much, no doubt, having been lost in translation. Plaintiff's counsel accepted, orally, that it was hard to work out just how the issue of publication of defamatory matter was aided by those decisions.

236 Having already noted the different kinds of defendants sued in defamation proceedings in connection with the internet, we must say something about those different classes of defendants and the decision or decisions pertaining to them.

237 In Bunt,[138] the defendants were persons who posted allegedly defamatory comments on a website, and the internet service providers which enabled the posts to be made. Justice Eady upheld an application by the ISPs that the proceeding against them be struck out on the basis that they were not the publishers of the posts. His Lordship described the role of ISPs, relevantly, as performing 'no more than a passive role in facilitating postings on the internet'.[139] Such a party, performing that role, should not be deemed to be a publisher at common law.[140] The position of a telephone company provided a comparable analogy. His Lordship contrasted the position of a passive conduit with that of a distributor of - archetypally - printed material.

238 Insofar as a difference might be discerned between those two situations, it may be said that a person in the chain of distribution knows what it is that is being distributed, albeit not knowing the defamatory matter lying within what is distributed. By contrast, an ISP provides a medium by which material can be conveyed between A and B, but - at least ordinarily - will have no knowledge whether any particular piece of material has been conveyed containing the allegedly defamatory matter.

239 In Trkulja (No 5), Beach J concluded that, if Bunt and two other English cases to which we will refer[141] were not distinguishable, and if it be thought that they would compel a conclusion that Google had not been a publisher of the defamatory matter as found by the jury, then Bunt and those other English decisions did 'not represent the common law of Australia'.[142]

240 Bunt has been criticised by Collins.[143] On his analysis, Bunt was the first case to draw a distinction at common law between mere conduits on the one hand and distributors on the other. The need to draw such a distinction was not obvious. Justice Eady did not determine that the Byrne line of authority would not apply, but rather that it did not apply because the ISPs were mere conduits.

241 Other academic writings have likewise expressed doubt as to the reasoning in Bunt.[144]

242 Despite the academic criticism, however, in Tamiz the Court of Appeal did not criticise the decision in Bunt.[145] So also, in Bleyer, having given consideration to Bunt and the other English cases, McCallum J differed from the conclusion of Beach J that there was any relevant difference between those decisions and the common law of Australia.[146] In that connection, her Honour referred, inter alia, to the observations of the Court of Appeal in Tamiz.[147]

243 If the 'mere conduit' approach be not accepted - a matter which need not be decided - our present view, in the case of an ISP which acts as a conduit between A and B, is that it should be regarded as a secondary publisher, but able to call in aid the common law or statutory defence of innocent dissemination,[148] this applying the Emmens line of authority. Such an analysis would seem preferable to the Byrne analysis. A factual difference underlies those lines of authority, and care ought be taken not to conflate them. The Emmens line concerns a party which intendedly participates in the chain of distribution of material that includes defamatory matter, but can establish, in effect, that its involvement was free of blame. The Byrne line, on the other hand, involves a party which has had no part in the distribution of the defamatory matter, but which has inadvertently provided a place - a wall or other space - which permits the publication of that matter by another. In some circumstances, post-notification, it is taken to become a primary publisher of that matter by adoption, acquiescence, or approval. Notice of the publication is relevant in each context, but it is for different reasons. In the former case, it can be relevant to the question whether a defendant which has to that point been an innocent disseminator has lost its protected position. In the latter case, it is the starting point for consideration whether the defendant adopted or approved the defamatory matter, and thereby became a primary publisher.

Blogs. Facebook. Bulletin boards. Forum hosts. YouTube.

244 The common features of websites, of a number of different kinds, which are established by the operator to permit posts being made by internet users is that the operator sets up the site and encourages its use, but that the content on the site is the work of those who make the posts. At least nominally, and perhaps in fact, the operator has the ability to see what has been posted and, at least after notice has been given of objectionable content, to take down that material. The juridical explanation given by the decisions to which we were referred has not been identical, but the decisions agree that the operator of such a site can or will be liable as publisher once notice has been given and a reasonable time to consider the notice and to take the post down has been allowed. That is either because the operator becomes a primary publisher at that time, or else because the operator then ceases to be an innocent secondary publisher. The first analysis was preferred by the Court of Appeal in Tamiz (where the defendant was the operator of the site blogger.com), and by the New Zealand Court of Appeal in Murray (where the defendant was a Facebook host). The second analysis was advanced by the Hong Kong Court of Final Appeal in Fevaworks (where the defendant was the host site of internet discussion forums).

245 Tamiz was an appeal from the decision of Eady J on an application to set aside leave to serve out of the jurisdiction. Google was sued in respect of blogs posted on a blogging platform which it provided. It had the ability to remove or block access to material once its attention was directed to that material. The plaintiff made complaint and the offending blogs were eventually removed. Although the plaintiff confined his case to the post-notification period, Eady J held that the defendant was not the publisher of the posts either before or after notification. The Court of Appeal sustained his Honour's conclusion with respect to publication before notification, but decided that the defendant could be a publisher post-notification, that being a matter of evidence.

246 The Court of Appeal's conclusions, described more fully, were as follows:

(1) Simply by providing the blogger service, the defendant did not become a primary publisher of the defamatory comments at common law.[149] It facilitated publication of the blog and the comments which were posted, but it had no prior knowledge and/or control over the content of the blog.

(2) It was very doubtful whether the defendant, before notification, was even a secondary publisher. But even if it was to be characterised as a disseminator, it should not be held to be a publisher before notification, since it neither knew nor, by the exercise of reasonable care, ought to have known that the blogs were likely to be defamatory.[150]

(3) But it would be different if it was found that it had permitted defamatory material to remain on the site after notice for too long a period. An inference of association with, or acquiescence in, the publication might be inferred. The defendant would be a publisher in those circumstances.[151]

247 In short, the preferable view, according to his Lordship, was that Google was neither a primary nor a secondary publisher before notification. But it might become a primary publisher, applying Byrne, after notification. The conception was that the provision of a platform for blogs was 'equivalent to the provision of a notice board.' Given notice, Google had the ability to remove a blog. There might be a finding, depending upon the evidence, that subsequent to notice the defendant had associated itself with, or made itself responsible for, the continued presence of the defamatory material on the blog, thereby becoming a publisher of that material.

248 There was no inherent contradiction within his Lordship's analysis. It was not a question of Google being an (innocent) secondary publisher before notice, and a possible primary publisher thereafter. Rather, Google was neither a primary nor a secondary publisher before notice, but might become a primary publisher thereafter.

249 Murray was an appeal by two of three defendants against the refusal of a judge to strike out the entirety of the statement of claim in a defamation proceeding brought against them by the plaintiff, Wishart.

250 It appears that the defendants became aware of the fact that the plaintiff was about to publish a book in connection with the murder of two infants some years earlier. The murders, and the subsequent trial and acquittal of the father, and the possible involvement of the children's mother in the murders, had been matters of controversy.

251 In the event, the first defendant established a Facebook page inviting boycott of the book which was soon to be published, and he used a Twitter account to publicise the Facebook page. He and the second defendant, and as well many others, posted comments on the Facebook page.

252 The plaintiff relied upon four causes of action. Relevantly, the causes of action pleaded against the first defendant involved allegedly defamatory statements made by him on his Twitter account and Facebook page. The cause of action against the second defendant rested on statements made by her on the first defendant's Facebook page.

253 In addition to the plaintiff's claims thus far identified, the plaintiff pleaded, as against the first defendant and a third defendant, Dimension Data New Zealand Limited ('DDNZ'), a claim based upon comments posted on the Facebook site by persons other than the first and second defendant. DDNZ was the first defendant's employer.

254 For present purposes, it is enough to say that, before the primary judge, the first defendant failed to make out his argument that he was not the publisher of the third party statements on the Facebook page, the judge concluding that there was an arguable case that he was the publisher of those statements.

255 The state of the evidence before the primary judge was that the first defendant had used the Facebook website to create a page. The operator of the site retained ownership of the server space and ultimate control over the contents of the page which he created. He, the first defendant, could not exercise 'meaningful editorial control' over comments before they were posted; but he could, once aware of comments published, retrospectively remove offensive material and block specific Facebook users from publishing further comments. He had done so in this case.

256 There was also some, although disputed, evidence that the plaintiff had warned the first defendant about defamatory postings.

257 Pausing, it can be seen that, unlike the situation in Tamiz and in Fevaworks (to the latter of which we will soon refer) the plaintiff did not sue the operator of the website which hosted (in this case) the first defendant's Facebook page. Rather, he sued the person who set up the page, that person's employer, and another person who posted comments on the first defendant's page.

258 The Court of Appeal allowed the appeal in respect of the third party statements, to the extent that it ordered that the existing relevant parts of the statement of claim were not tenable. Further, it 'limit[ed] any renewed claims in relation to third party statement to claims that the first [defendant] knew of any such statement and failed to remove it within a reasonable time in circumstances that give rise to an inference that he was taking full responsibility for such statement'.[152]

The mere fact that a publication occurs on the internet does not provide a basis for some generalised rule as to the definition of a publisher. The position of an ISP, for example, is materially different from that of the person who creates a Facebook page as [the first defendant] did here.[153]

260 The Court next observed that analysis of authorities required it to 'apply reasoning by strained analogy, because the old cases [did not] deal with publication on the internet'.

261 Having exhaustively reviewed authorities, the Court emphasised

how sensitive the outcome can be to the particular circumstances of the publication. The fact that many of the authorities relate to publication in one form or another on the internet does not provide any form of common theme, because of the different roles taken by the alleged publisher in each case.[154]

262 The Court did not regard the situation of the news vendor in Emmens as a particularly apposite analogy with the host of the Facebook page in the present case. It observed that -

The news vendor is a publisher only because of the role taken in distributing the primary vehicle of publication, the newspaper itself. This contrasts with the host of a Facebook page which is providing the actual medium of publication, and whose role in the publication is completed before publication occurs.[155]

263 Acknowledging that the analogy was not perfect, the Court found the more closely analogous situation to be that which arose in Byrne. Noting that the Court of Final Appeal of Hong Kong, in Fevaworks, had rejected the analogy on the ground that posting a notice on the wall in Byrne was a breach of the rules of the club and therefore amounted to a trespass - by contrast with posting a message on a Facebook page in response to an invitation to do so - the Court said that it did

not consider the fact that the posting of a notice on the wall in Byrne v Deane was a breach of the club's rules was a factor affecting the outcome in that case. The decisive factor was that the club and its owners had not posted the defamatory notice and, until they became aware of it, were in no position to prevent or bring to an end the publication of the defamatory message.[156]

264 The 'ought to know' test was propounded in Emmens in a context which the Court of Appeal dismissed as being not a useful analogy. The Court nonetheless considered whether an 'ought to know' test should somehow be attached to consideration of the potential liability of the creator of a Facebook page. Having considered a number of concerns in that connection, the Court concluded that -

These concerns lead us to conclude that the actual knowledge test should be the only test to determine whether a Facebook page host is a publisher. That is consistent with at least some of the authorities to which we have referred, (Sadiq v Baycorp (NZ) Ltd, A v Google New Zealand Ltd and Davison v Habeeb) and with the Law Commission's analysis. It conforms with the approach in Byrne v Deane, which is, we believe, the most appropriate analogy and with the decision in Urbanchich v Drummoyne Municipal Council. It makes the liability risk of a Facebook page host no greater than that of an organiser of a public meeting - another appropriate analogy, in our view. It is consistent with the right of freedom expression in the Bill of Rights, bearing in mind the unavailability of the innocent dissemination defence. And it provides a situation where liability for defamation is not imposed on the basis of negligence.[157]

In Murray v Wishart,[158] the New Zealand Court of Appeal held that an actual knowledge test and not a constructive knowledge test should be applied to determine whether a website forum host - in that case a Facebook page host - is a secondary publisher in respect of third party postings.[159] As a matter of principle, I do not consider that a different test should apply to a publisher of internet material from that applying to a publisher of physical, broadcast or televised material. In any event, as a matter of authority the approach of the New Zealand Court of Appeal is not open in Australia given the High Court's endorsement of the constructive knowledge test in the case of secondary participants.[160]

266 His Honour was indisputably correct in saying that, in Australia, so far as a secondary publisher is concerned, the constructive knowledge test has a part to play. To the extent that Murray said the contrary, it could not be accepted. But in Murray, the Court of Appeal rejected the idea that the first defendant could be viewed as a secondary publisher. Rather, the Court's analysis was that the first defendant's possible liability was, on a Byrne analysis, as a primary publisher after notification.

267 In Fevaworks, the plaintiffs brought actions for libel against the defendants, who were the providers, administrators and managers of a website which hosted a popular internet discussion forum in Hong Kong. The actions were in respect of three batches of defamatory statements posted by certain users of the forum. At trial, the judge had dismissed the plaintiffs' claims in respect of defamatory statements posted on 24 October 2008, of which the defendants were informed by the plaintiffs on 27 October 2008, and which were removed within a few hours. The trial judge also dismissed a claim in respect of postings made on 21 January 2009, discovered by the defendants some 12 hours after they had been posted, and thereupon immediately removed. It was not in dispute that the defendants were unaware of the 2008 postings until informed of them by the plaintiffs, and were unaware of the existence of the 2009 postings until they themselves discovered those postings.

268 The trial judge's decision was upheld by the Court of Appeal. Leave to appeal was granted in respect of this question:

In respect of statements defamatory of a third party posted on a commercial website which may be and were accessed by another party, whether the defence of innocent dissemination or any other defence is available to a commercial website host, and if so, under what circumstances may such defence be established or defeated.

269 Permanent Judge Ribeiro, who gave the main judgment, said that 'one of the main issues dividing the parties is whether the respondents are eligible to rely on the innocent dissemination defence or whether they are the first or main publishers to whom it does not apply'.[161] Permanent Judge Ribeiro differentiated the Emmens and Byrne lines of authority this way:

1. It is important to appreciate the differences between the principles which derive from the notice board line of cases and those which constitute the innocent dissemination defence.
2. As analysed above, the innocent dissemination defence is a common law doctrine developed to mitigate the harshness of the strict publication rule. While it does not avail the first or main publishers it brings relief to subordinate publishers who have knowingly participated in the process of disseminating the article concerned. The defence is therefore applicable to persons who are admittedly publishers, although playing a subordinate role, allowing them to be exonerated from liability if they discharge the burden of showing that they did not know that the article which they had helped to disseminate contained the offending content and that such lack of knowledge was not due to any lack of reasonable care on their part.
3. In contrast, the occupiers in the notice board and graffiti cases have not in any sense assisted or participated in the originator's publication of the libel. These are not cases involving anyone who has played a role in a scheme for distributing the offending publication seeking relief from the strict rule. As Greene LJ pointed out in Byrne v Deane, the posting of the defamatory verse on the club's notice board was an act of trespass. A person who defaces an occupier's walls with graffiti prima facie does not do so with the approval or encouragement of the occupier. What this line of cases addresses are the conditions which must be satisfied before the occupier can be regarded as having turned himself into a publisher of the libel. That occurs only if the plaintiff shows that the occupier became aware of the libellous statement on his premises and allowed it to remain in place in circumstances which justify the inference that he has adopted or ratified its publication. If the occupier was not aware or if the inference cannot properly be drawn, he does not make himself a publisher of the trespassing inscription and is not liable at all. These rules are therefore not about relieving a person's prima facie liability as a publisher but about whether or not an occupier is constituted a publisher. Since it is about establishing the occupier's liability as a publisher (and not about a publisher establishing a defence), the onus rests on the plaintiff.
4. It follows from the distinctions drawn above that I do not share the approach in some of the reported cases involving libel in an internet context where the Byrne v Deane principles appear to have been treated merely as a facet of the innocent dissemination defence or the defence under section 1 of the Defamation Act 1996, in the United Kingdom.[162]

270 Permanent Judge Ribeiro concluded that the Byrne principles were inapplicable:

1. Once the nature of the Byrne v Deane principles is grasped, it becomes clear that they do not apply to internet platform providers like the respondents. The provider of a discussion forum is in a wholly different position from that of the occupier of premises who is not in the business of publishing or facilitating publication at all, but who has had imposed on him the defamatory act of a trespasser.
2. The respondents plainly played an active role in encouraging and facilitating the multitude of internet postings by members of their forum. As described in Section B of this judgment, they designed the forum with its various channels catering for their users' different interests; they laid down conditions for becoming a member and being permitted to make postings; they provided users browsing their website access to the discussion threads developed on their forum; they employed administrators whose job was to monitor discussions and to delete postings which broke the rules; and they derived income from advertisements placed on their website, a business model which obviously benefits from attracting as many users as possible to the forum.
3. The respondents were therefore, in my view, plainly participants in the publication of postings by the forum's users and in that sense they were publishers from the outset, it being in issue whether they were first or main publishers or merely subordinate publishers. I accept Mr Thomas SC's submission that they were in a substantively different position from the occupiers in the notice board and graffiti cases. The relevant question in the present case is whether, as publishers, the respondents are entitled to rely on, and have established, the defence of innocent dissemination, relieving themselves of the strict publication rule which would otherwise be applicable. The question is not whether, originally being non-publishers, they have, when fixed with knowledge of the defamatory postings, demonstrated their consent to and adoption of those postings, turning themselves into publishers.
4. In this context, I respectfully part company with the reasoning (adopted on an interlocutory basis) of the English Court of Appeal in Tamiz v Google Inc. It is reasoning which proceeds on the basis that successful invocation of the defence of innocent dissemination results in the defendant being deemed not to have published at all. For the reasons previously given, I do not accept that premise. Nor am I able to accept the distinction drawn between the notice board and graffiti analogies, nor the suggestion that "the provision of a platform for blogs is equivalent to the provision of a notice board". As indicated above, my view is that the provider of an internet discussion platform similar to that provided by the respondents falls from the outset within the broad traditional concept of "a publisher", a characteristic not shared by a golf club or other occupier who puts up a notice board on which a trespassing message is posted.
5. I pause to note in passing that there is rightly no suggestion in the present case that the providers of a discussion platform like the respondents should be regarded as "mere conduits" or "passive facilitators" comparable to the Post Office or a telephone company, as has sometimes been argued in certain internet cases.[163]

271 That conclusion, as his Honour recognised, was contrary to the position earlier expressed by the Court of Appeal in Tamiz.[164] It was also inconsistent with the later conclusion of the New Zealand Court of Appeal in Murray.

272 The conclusions that Byrne principles were inapplicable, and that the respondents were publishers from the outset led to focus being placed upon a question whether they were primary or secondary publishers of the posted comments. If it were the latter, then an innocent dissemination defence would be potentially available at common law.

273 Permanent Judge Ribeiro rejected a submission for the appellants that the respondents should be taken to have authorised the defamatory publications, and so should be treated as primary publishers. The exchanges on the platform which they provided contraindicated such a conclusion.

274 His Honour identified these characteristics of a primary publisher:

(i) that he knows or can easily acquire knowledge of the content of the article being published (although not necessarily of its defamatory nature as a matter of law); and (ii) that he has a realistic ability to control publication of such content, in other words, editorial control involving the ability and opportunity to prevent publication of such content. [165]

275 As to the knowledge of content required of a primary publisher, Ribeiro PJ adopted what had been said by Eady J in Bunt:

The important question is whether the publisher knew or can properly be expected to have known the content of the article being published. Eady J stated that knowledge of "the relevant words" contained in the article complained of must be shown. That should be taken to mean that the publisher must know or be taken to know the content - not necessarily every single word posted - but the gist or substantive content of what is being published, to qualify as a first or main publisher. Such knowledge may exist in relation to the content of a particular posting or a particular discussion thread or group of discussion threads, it being irrelevant whether the provider realised that such content was in law defamatory. I reject in any event the appellants' suggestion that a discussion forum provider should be treated as having knowledge of the content of every message posted on the forum and deemed to be a first or main publisher thereof.[166]

276 Having dealt next with the 'control criterion', Ribeiro PJ expressed this conclusion:

When the abovementioned criteria are applied to the respondents, it is in my view clear that they are subordinate publishers and not first or main publishers of the defamatory postings. They were certainly publishers of those postings (and do not seek to argue otherwise) since they provided the platform for their dissemination, but the respondents were not aware of their content and realistically, in a many-to-many context, did not have the ability or opportunity to prevent their dissemination, having learned of them only after they had already been published by their originators. The respondents are therefore in principle able to invoke the innocent dissemination defence as subordinate publishers.[167]

277 As his Honour pointed out, the focus of the innocent dissemination defence in the past had been upon completed publication. Thus, for instance, the newsagent who had already sold a consignment of newspapers. But assuming that the originator of the platform which permitted Facebook postings to be made was a secondary publisher, was a distinction to be drawn between the circumstances of past and continuing publication? His Honour's answer was this:

In my view, it is consistent with the policy underpinning the defence that the same standard of reasonableness should be applied in a situation of acquired knowledge. A subordinate publisher should be afforded the continued protection of the defence if he proves that upon becoming aware of such content, he promptly took all reasonable steps to remove the offending content from circulation as soon as reasonably practicable.[168]

278 The appellate decisions in Tamiz, Murray and Fevaworks all post-dated the decision of Moreland J in Godfrey v Demon Internet Pty Ltd.[169] There, his Lordship refused an application by the plaintiff to strike out the defence of a forum host denying publication of posts on the site. The forum host was an ISP, the forum hosting facility being an 'add-on' to the defendant's ordinary activity. The case was advanced by the plaintiff as one of publication after notification. His Lordship's decision, however, travelled beyond what was strictly necessary to decide the application. The rationale of his decision was that by storing posts, and then transmitting them to an enquirer, it was arguable that the defendant published the posts - though, it appears, as a secondary publisher. The circumstances were different to those considered in Bunt, because there the ISPs did not host any website relevant to the claims, but merely afforded connection to the internet, this facilitating posting to a message board. Justice Eady distinguished Godfrey in Bunt.

279 Godfrey was referred to by the Court of Appeal in Tamiz. Lord Justice Richards considered that Davison v Habeeb[170] was more in point.

280 Godfrey was also referred to by O'Regan P and Ellen France J in Murray.[171] The Court took the decision to have been based on the defendant having been a secondary publisher.

281 Davison, cited by the Court of Appeal in Tamiz, was a case in which Google was sued as the operator of a blogging host site. Judge Park QC, sitting as a deputy judge of the Queen's Bench division, concluded that Google was arguably liable as a secondary publisher from the outset - but protected by an innocent dissemination defence until put on notice; or else potentially liable after being put on notice in accordance with the Byrne analysis. Either approach led to the same potential liability.

282 For completeness, we mention the recent decision of Horner J in Galloway v Frazer,[172] which concerned video posted on a YouTube facility provided by Google. There, following the Court of Appeal decision in Tamiz, his Lordship adopted the Byrne analysis.[173]

283 As we earlier noted, there is a disagreement in the authorities whether a forum provider, upon whose site internet users post material, is akin to a person in a chain of distribution which begins with creation of the allegedly defamatory material and continues with its distribution from one entity to the next - in which case the provider will be a secondary publisher for whom an innocent dissemination defence will potentially be available; or whether the better view is that the operator of such a site may be liable as publisher of a posted comment subsequent to notification and after a reasonable time has elapsed for removal of the offending post, akin to the situation in Byrne. Either way, in practical terms, the liability or otherwise of a defendant in a particular case is unlikely to differ. Nonetheless, there are conceptual differences in the two analyses.

284 It is not necessary for us to express a concluded view as to the preferable analysis. It is sufficient to say that each analysis has its own difficulties; but that, at least in most instances. the practical consequences for the parties of the application of either analysis would probably be the same.[174]

285 Whatever doubt there may be about aspects of defamation law to do with the liability of site operators of the kind which we have been discussing, we think it is very clear that the circumstances there in point differ significantly from the circumstances which arise in connection with search engines. In the forum host cases, the host provides the site and the means of utilising it. To a greater or lesser extent, the site operator has the ability to control the continued presence of objectionable material on its site. Its purpose is to make the site available for comments by others. In the case of a search engine provider, however, the alleged publisher has no connection, other than by operation of algorithms, with the websites, created by others, which contain matter that the search engine crawls, indexes, and then returns as results in response to a search term. Moreover, the search engine provider has no role at all in the selection of the search term which is inputted by an internet user. That is not less the case because the autocomplete or related searches functions, for reasons which we have previously explained, may suggest completions of the search terms. In the case of a search engine, there is no possible analogy, as we perceive it, with the conception of an electronic notice board, or a golf club wall, or the wall of a bus shelter. So decisions respecting the liability of website hosts sued as publishers of posted comments, in our opinion, cannot simply be transposed to the search engine situations. We respectfully agree with the observations of the New Zealand Court of Appeal in Murray, warning against the idea of there being a 'one size fits all' analysis of publication in the context of the internet.

286 We referred a moment ago to 'search engine situations'. Use of the plural was intended because the decided cases show that there are multiple variants of claims against search engine providers.

287 The search engine decisions which were drawn to our attention were Metropolitan Schools, Crookes, Trkulja (No 5), Bleyer, A v Google New Zealand,[175] Yeung, Duffy and Niemela v Malamas[176]. We will address them in that order.

288 In Metropolitan Schools, the plaintiff's claim against Google was that when its search engine was used to search for adult learning courses provided by the plaintiff, the search results included links to allegedly defamatory comments posted on a forum provided by the first defendant. Justice Eady upheld Google's application to set aside leave to serve out of a jurisdiction on the basis that it was not reasonably arguable that this defendant was the publisher of the material complained of.

289 His Lordship recognised that the circumstances and characteristics of defendants in earlier decisions - Godfrey and Bunt - were in some respects different from the circumstances and characteristics of a search engine defendant; but he concluded that they were not materially different.

The appropriate question here, perhaps, is whether the third defendant should be regarded as a mere facilitator in respect of the publication of the "snippet" and whether, in particular, that would remain a proper interpretation even after the date of notification.[177]

It has been recognised, at common law, that for a person to be fixed with responsibility for publishing defamatory words, there needs to be present a mental element. I summarised the position in Bunt v Tilley:

[21] In determining responsibility for publication in the context of the law of defamation, it seems to me to be important to focus on what the person did, or failed to do, in the chain of communication. It is clear that the state of a defendant's knowledge can be an important factor. If a person knowingly permits another to communicate information which is defamatory, when there would be an opportunity to prevent the publication, there would seem to be no reason in principle why liability should not accrue. So too, if the true position were that the applicants had been (in the claimant's words) responsible for 'corporate sponsorship and approval of their illegal activities'.

[22] I have little doubt, however, that to impose legal responsibility upon anyone under the common law for the publication of words it is essential to demonstrate a degree of awareness or at least an assumption of general responsibility, such as has long been recognised in the context of editorial responsibility. As Lord Morris commented in McLeod v St Aubyn [1899] AC 549 at 562:

'A printer and publisher intends to publish, and so intending cannot plead as a justification that he did not know the contents. The appellant in this case never intended to publish.'

In that case the relevant publication consisted in handing over an unread copy of a newspaper for return the following day. It was held that there was no sufficient degree of awareness or intention to impose legal responsibility for that 'publication'.

[23] Of course, to be liable for a defamatory publication it is not always necessary to be aware of the defamatory content, still less of its legal significance. Editors and publishers are often fixed with responsibility notwithstanding such lack of knowledge. On the other hand, for a person to be held responsible there must be knowing involvement in the process of publication of the relevant words. It is not enough that a person merely plays a passive instrumental role in the process.[178]

1. When a search is carried out by a web user via the Google search engine it is clear, from what I have said already about its function, that there is no human input from the third defendant. None of its officers or employees takes any part in the search. It is performed automatically in accordance with computer programs.
2. When a snippet is thrown up on the user's screen in response to his search, it points him in the direction of an entry somewhere on the web that corresponds, to a greater or lesser extent, to the search terms he has typed in. It is for him to access or not, as he chooses. It is fundamentally important to have in mind that the third defendant has no role to play in formulating the search terms. Accordingly, it could not prevent the snippet appearing in response to the user's request unless it has taken some positive step in advance. There being no input from the third defendant, therefore, on the scenario I have so far posited, it cannot be characterised as a publisher at common law. It has not authorised or caused the snippet to appear on the user's screen in any meaningful sense. It has merely, by the provision of its search service, played the role of a facilitator.
3. Analogies are not always helpful, but there will often be resort to analogy when the common law has to be applied to new and unfamiliar concepts. Here, an analogy may be drawn perhaps with a search carried out in a large conventional library. If a scholar wishes to check for references to his research topic, he may well consult the library catalogue. On doing so, he may find that there are some potentially relevant books in one of the bays and make his way there to see whether he can make use of the content. It is hardly realistic to attribute responsibility for the content of those books to the compiler(s) of the catalogue. On the other hand, if the compilers have made an effort to be more informative, by quoting brief snippets from the book, the position may be different. Suppose the catalogue records that a particular book contains allegations of corruption against a living politician, or perhaps it goes further and spells out a particular activity, such as "flipping" homes to avoid capital gains tax, then there could be legal liability on the part of the compiler under the "repetition rule": see eg Gatley on Libel and Slander (11th edn, 2008) paras 11.4 and 32.8.
4. No doubt it would be said here too, by analogy, that the third defendant should be liable for repeating the "scam" allegations against the claimant. Yet, whereas a compiler of a conventional library catalogue will consciously at some point have chosen the wording of any "snippet" or summary included, that is not so in the case of a search engine. There will have been no intervention on the part of any human agent. It has all been done by the web-crawling "robots".
5. The next question is whether the legal position is, or should be, any different once the third defendant has been informed of the defamatory content of a "snippet" thrown up by the search engine. In the circumstances before Morland J in Godfrey v Demon Internet Ltd [1999] EWHC QB 244; [1999] 4 All ER 342, [2001] QB 201, the acquisition of knowledge was clearly regarded as critical. That is largely because the law recognises that a person can become liable for the publication of a libel by acquiescence; that is to say, by permitting publication to continue when he or she has the power to prevent it. As I have said, someone hosting a website will generally be able to remove material that is legally objectionable. If this is not done, then there may be liability on the basis of authorisation or acquiescence.
6. A search engine, however, is a different kind of internet intermediary. It is not possible to draw a complete analogy with a website host. One cannot merely press a button to ensure that the offending words will never reappear on a Google search snippet: there is no control over the search terms typed in by future users. If the words are thrown up in response to a future search, it would by no means follow that the third defendant has authorised or acquiesced in that process.
7. There are some steps that the third defendant can take and they have been explored in evidence in the context of what has been described as its "take down" policy. There is a degree of international recognition that the operators of search engines should put in place such a system (which could obviously either be on a voluntary basis or put upon a statutory footing) to take account of legitimate complaints about legally objectionable material. It is by no means easy to arrive at an overall conclusion that is satisfactory from all points of view. In particular, the material may be objectionable under the domestic law of one jurisdiction while being regarded as legitimate in others.
8. In this case, the evidence shows that Google has taken steps to ensure that certain identified URLs are blocked, in the sense that when web-crawling takes place, the content of such URLs will not be displayed in response to Google searches carried out on Google.co.uk. This has now happened in relation to the "scam" material on many occasions. But I am told that the third defendant needs to have specific URLs identified and is not in a position to put in place a more effective block on the specific words complained of without, at the same time, blocking a huge amount of other material which might contain some of the individual words comprising the offending snippet.
9. It may well be that the third defendant's "notice and take down" procedure has not operated as rapidly as Mr Browne and his client would wish, but it does not follow as a matter of law that between notification and "take down" the third defendant becomes or remains liable as a publisher of the offending material. While efforts are being made to achieve a "take down" in relation to a particular URL, it is hardly possible to fix the third defendant with liability on the basis of authorisation, approval or acquiescence.
10. These practical difficulties also impact upon the feasibility and effectiveness of injunctive relief. That is obviously a separate issue from that of responsibility for publication, but it is another illustration of the powerlessness of the third defendant to control what is thrown up by Google searches.[179]

Against this background, including the steps so far taken by the third defendant to block the identified URLs, I believe it is unrealistic to attribute responsibility for publication to the third defendant, whether on the basis of authorship or acquiescence. There is no doubt room for debate as to what further blocking steps it would be open for it to take, or how effective they might be, but that does not seem to me to affect my overall conclusion on liability. This decision is quite independent of any defence provided by s 1(1) of the 1996 Act, since if a person is not properly to be categorised as the publisher at common law, there is no need of a defence: see eg Bunt v Tilley [2006] EWHC (QB) 407; [2006] 3 All ER 336 at [37], [2007] 1 WLR 1243.[180]

292 His Lordship thus concluded, in essence, that Google was not a publisher of the search result 'snippets' at the outset; and did not later become the publisher by failing to remove all possible links to the sites which produced the snippets. Although his Lordship left open the possibility that, on the evidence, authorisation or acquiescence might be found in a particular case, the position was not to be equated with that of a website host. Particularly, as we understand it, his Lordship perceived that a web search response reflects only a moment in time, whereas a posted comment is present until removed.

293 The language of 'authorisation or acquiescence' was evidently not a reference to the Byrne line of authority. His Lordship made no reference to Byrne at all. It is tolerably clear that his position was that if a search engine proprietor was to be a publisher at all, it would be as a secondary publisher. That would give rise to the question whether the defendant could (successfully) rely upon an innocent dissemination defence. As his Lordship perceived it, the defendant might very well fail to bring itself within s 1 of the Defamation Act 1996 (UK) - which is considerably different to s 32 of the Defamation Act 2005.

294 In Crookes, the plaintiff relied upon 'publication' of hyperlinks to sites containing allegedly defamatory matter. On the site which provided the hyperlinks, there was no comment as to the content of the hyperlinked pages. The Supreme Court of Canada upheld the decision of the trial judge that hyperlinks, without more, were akin to footnotes, and did not republish the content of the hyperlinked pages. In different ways, however, the judgments reserved the position as to what would have been the consequence if defamatory content had been attached to the identification of the hyperlinks.

295 Pausing, in this case, as we have earlier noted, the plaintiff makes no attempt to rely upon the content of hyperlinked pages, or upon the content of pages that could be identified by clicking on a thumbnail. The alleged publication and defamatory matter begins and ends with the images, web matter and autocomplete predictions.

296 In Trkulja (No 5), the plaintiff sued upon images and web search results, and upon an underlying page on a hyperlinked site.[181] Justice Beach refused a non obstante veredicto application, the jury having found that - (1) Google had published the search results and the underlying article; (2) one defamatory imputation was conveyed; (3) a defence of innocent dissemination was made out for most but not all of the period of alleged publication of the images matter;[182] and (4) innocent dissemination was made out for the entirety of the period of alleged publication of the web matter.[183]

297 At the heart of the reasoning in Trkulja (No 5) were the following passages, cited in full by the judge below:

1. The question of whether or not Google Inc was a publisher is a matter of mixed fact and law. In my view, it was open to the jury to find the facts in this proceeding in such a way as to entitle the jury to conclude that Google Inc was a publisher even before it had any notice from anybody acting on behalf of the plaintiff. The jury were entitled to conclude that Google Inc intended to publish the material that its automated systems produced, because that was what they were designed to do upon a search request being typed into one of Google Inc's search products. In that sense, Google Inc is like the newsagent that sells a newspaper containing a defamatory article. While there might be no specific intention to publish defamatory material, there is a relevant intention by the newsagent to publish the newspaper for the purposes of the law of defamation.
2. By parity of reasoning, those who operate libraries have sometimes been held to be publishers for the purposes of defamation law. That said, newsagents, librarians and the like usually avoid liability for defamation because of their ability to avail themselves of the defence of innocent dissemination (a defence which Google Inc was able to avail itself of for publications of the images matter prior to 11 October 2009, and all of the publications of the web matter that were the subject of this proceeding).

1. While much was made by Google Inc in the present case of Eady J's statements in Bunt and Tamiz that an internet service provider who performs no more than a passive role cannot be a publisher, those statements have to be seen in the light of the facts in those cases. To say as a general principle that if an entity's role is a passive one then it cannot be a publisher, would cut across principles which have formed the basis for liability in the newsagent/library type cases and also in those cases where someone with power to remove a defamatory publication chooses not to do so in circumstances where an inference of consent can be drawn.[184]
2. In any event, and putting to one side the factual differences I have identified, to the extent that there is anything written in the judgments of Bunt v Tilley,[185] Metropolitan Schools Ltd v Designtechnica Corporation[186] and Tamiz v Google Inc[187] that might be thought to compel the conclusion that on the facts of the present case it was not open to the jury to conclude that Google Inc was a publisher of either the images matter or the web matter, then the same does not represent the common law of Australia.[188] Further, while on the facts in Bunt, the defendants were correctly described as "internet intermediaries" (whatever may be the legal consequences of such a description), it is, with respect, doubtful that that same description can be applied to an internet search engine provider in respect of material produced as a result of the operation of that search engine. That said, any such "internet intermediary" is, in any event, performing more than the "merely passive role ... [of] facilitating postings" (Cf Bunt).
3. It follows that, in my view, it was open to the jury to conclude that Google Inc was a publisher - even if it did not have notice of the content of the material about which complaint was made. Google Inc's submission to the contrary must be rejected. However, Google Inc goes further and asserts that even with notice, it is not capable of being liable as a publisher "because no proper inference about Google Inc adopting or accepting responsibility complained of can ever be drawn from Google Inc's conduct in operating a search engine".
4. This submission must also be rejected. The question is whether, after relevant notice, the failure of an entity with the power to stop publication and which fails to stop publication after a reasonable time, is capable of leading to an inference that that entity consents to the publication. Such an inference is clearly capable of being drawn in the right circumstances (including the circumstances of this case). Further, if that inference is drawn then the trier of fact is entitled (but not bound) to conclude that the relevant entity is a publisher.[189] Google Inc's submission on this issue must be rejected for a number of reasons, the least of which is that it understates the ways in which a person may be held liable as a publisher.[190]

298 The questions put to the jury in that case assumed that Google was, if a publisher at all, then a secondary publisher of the search results and the underlying webpage. The ruling of Beach J, with its analogy between the Google search engine and a newspaper seller or librarian, was consistent with that analysis. However, his Honour's footnoted reference to Urbanchich was a little puzzling. For, as we have earlier noted, the rationale of that case - whereby a defendant can become a primary publisher - is not related to the position of a secondary publisher deprived of an innocent dissemination defence.

299 We mention two other matters at this stage. First, his Honour's analysis did not rely upon Thompson. Indeed, his Honour did not mention that authority at all. Second, we leave for later consideration his Honour's disagreement with Bunt, Metropolitan Schools and Tamiz.

300 The ruling of Beach J has been referred to in subsequent decisions - Bleyer, Tamiz (in the Court of Appeal),[191] Yeung, Murray and Niemela - but not with either uniform endorsement or disapproval.

301 A v Google NZ is the next decision chronologically. The plaintiff sued on snippets and a hyperlink. There were cross-applications for summary judgment. The plaintiff's application failed, and the defendant's succeeded. It did so on the footing that the plaintiff had sued the wrong defendant. Associate Justice Abbott also considered Google's 'non publication' argument. He held that the question raised was a novel one and that summary disposition would be inappropriate. He ventured the suggestion that a secondary publisher solution might be appropriate.[192]

302 In Bleyer, the plaintiff sued upon snippets and underlying webpages. Justice McCallum granted the defendant's application to permanently stay the proceeding. It was sufficient for her Honour to hold, and she did, that the resources which would be demanded of the Court and the parties were the matter to proceed to trial would be out of all proportion to the interest at stake.

303 But her Honour also considered the question whether the defendant could possibly be held to be the publisher of the snippets and underlying webpages. She analysed the decisions in Bunt, Metropolitan Schools and Tamiz[193] and addressed the conclusions of Beach J that - (1) it had been open to the jury to find intention to publish because the human-designed system operated as it was designed to do; and (2) insofar as Bunt, Metropolitan Schools and Tamiz stood to the contrary, they did not represent the common law of Australia.[194] She disagreed with both those conclusions, saying:

1. Importantly, however, his Honour further held that to the extent that the decisions in Bunt, Metropolitan International Schools and Tamiz might be thought to compel a different conclusion, they did not represent the common law of Australia: at [29]. In support of that proposition, his Honour cited the decision of Hunt J in Urbanchich v Drummoyne Municipal Council (unreported, NSWSC, Hunt J, 22 December 1988, BC8801175) (Urbanchich).
2. To the extent that the claim in Trkulja alleged that Google Inc was liable for publication before being put on notice of the plaintiff's claim, I would respectfully not share his Honour's view that the decision in Urbanchich was authoritative in that context. Urbanchich was a case which turned on the issue of notice. The proceeding before Hunt J was the separate trial of the issue whether the plaintiff could establish publication as against the second defendant, the Urban Transit Authority. The defendant council was sued in respect of posters placed on bus shelters under its control. The basis on which it was asserted that the authority was liable as a publisher was its failure to remove the posters after receiving notice of the plaintiff's complaint. The proposition for which the decision stands as authority, in my view, is that set out in the following passage of the judgment of Hunt J:

In a case where the plaintiff seeks to make the defendant responsible for the publication of someone else's defamatory statement which is physically attached to the defendant's property, he must establish more than mere knowledge on the part of the defendant of the existence of that statement and the opportunity to remove it. According to the authorities, the plaintiff must establish that the defendant consented to, or approved of, or adopted, or promoted, or on some way ratified, the continued presence of that statement on his property so that persons other than the plaintiff may continue to read it - in other words, the plaintiff must establish in one way or another an acceptance by the defendant of a responsibility for the continued publication of that statement.[195]

The evidence before me establishes that there is no human input in the application of the Google search engine apart from the creation of the algorithm. I would respectfully disagree with the conclusion reached by Beach J in Trkulja that the performance of the function of the algorithm in that circumstance is capable of establishing liability as a publisher at common law. I would adopt the English line of authority to the effect that, at least prior to notification of a complaint (and on the strength of the evidence before me), Google Inc cannot be liable as a publisher of the results produced by its search engine.[196]

304 Having held that the defendant was at least not liable as a publisher before notification, her Honour addressed the situation if that conclusion was incorrect. She held that the 'overwhelming likelihood' was that an innocent dissemination defence would succeed at least before notification. After notification, the issue was probably not amenable to summary determination.[197]

305 The inevitable inference flowing from her Honour's reference to an innocent dissemination defence is that, but for her conclusion that the defendant was not a publisher at all, she would have held that it was a secondary publisher. Only if such an assumption be made would notification be relevant.

306 In Yeung, the plaintiff sued only upon autocomplete and related search predictions. Deputy Judge Ng refused an application by Google to set aside leave to serve out of the jurisdiction, holding that there was a good arguable case that the defendant was a publisher, and more than a mere facilitator. She relied upon Fevaworks (which, as we have said, was not a search engine case) and Trkulja (No 5).[198] Her Honour extended this analysis to autocomplete predictions and related searches - which were the subject-matter of that proceeding - although she recognised that they were different to web search returns. Her Honour's conclusion that the defendant was arguably a publisher of the impugned material was evidently founded on it being a secondary publisher.

307 In Duffy, the plaintiff relied upon - (1) URLs and snippets; (2) underlying webpages; and (3) autocomplete predictions. The plaintiff had notified the defendant of the allegedly defamatory material on the webpages. Those pages were, in time, removed from search returns. The plaintiff also complained about an autocomplete prediction. The defendant did not take it down.

308 The plaintiff's claim was conducted entirely on the basis that the defendant was liable as publisher in the post-notification period. Justice Blue held that the conception of a secondary publisher is part of Australian law, and he described its content.[199] He rejected a submission for Google that a defendant can only ever be a publisher if it authorises or accepts responsibility for the publication. That was a misunderstanding of the reach of Urbanchich.[200]

309 His Honour then addressed the issue of publication in the context of the internet. Having considered the decisions to which we have referred, his Honour expressed these conclusions:

1. The first issue that arises is whether the paragraphs (title, snippet and URL) displayed by the Google websites to users in response to searches for Dr Duffy's name were published by Google. I take as an exemplar for this purpose the first paragraph extracted at [11] above, namely:

R1 Ripoff Report Janice Duffy - Psychic Stalker Psychics Beware Of...

Dr Janice Duffy is truly an embarrassment to her profession as a Senior Researcher in Adelaide Australia #2 Consumer Comment. Respond to this report...

www.ripoffreport.com/...Janice-Duffy.../janice-duffy-psychic-stalker-98d93.htm Cached

1. Google was the sole operator and controller of the Google website. The paragraphs resided on Google's website. The paragraphs were communicated by Google to the user conducting a search. Google played a critical role in communicating the paragraphs to the user. The physical element of publication is present. Google did not play the passive role of a mere conduit such as an internet service provider who merely provides access to the internet or a telecommunications carrier who merely provides access to the telephone network. Google played an active role in generating the paragraphs and communicating them to the user. The mere fact that the words are programmed to be generated because they appear on third party webpages makes no difference to the physical element. It makes no difference to the physical element whether a person directly composes the words in question or programs a machine which does so as a result of the program. I agree with the analysis of Beach J in Trkulja v Google Inc LLC (No 5) in this respect.
2. As to the mental element, Google intended to publish the paragraphs to the user. In addition, because Google was not a primary publisher, it is necessary that Google knew or ought to have known of the existence of the paragraphs. Google had no such knowledge before Dr Duffy's notifications in or in some cases after September 2009 and cannot be regarded as a publisher of the paragraphs before that time. However, if Google acquired knowledge of the paragraphs by reason of Dr Duffy's notifications and failed to remove them within a reasonable time thereafter, the necessary mental element will be present.
3. The mere fact that the paragraphs were generated automatically by Google's software programs does not prevent Google being a publisher of them after notification by Dr Duffy. If Google personnel were made aware of the existence of the paragraphs generated by Google's own software programs and failed to remove them, their continuing existence thereafter was the direct result of human action or inaction rather than merely the result of machine operation.
4. Approaching the question from first principles, Google was a secondary publisher of the paragraphs after notification and lapse of a reasonable time to allow for their removal (if that occurred).
5. There is authority which supports this conclusion. In Trkulja v Google Inc LLC (No 5), Beach J held that it was open to the jury to conclude that Google published the paragraphs in that case after notification. While this was a ruling on a non-obstante application, there were no evidentiary or factual issues for the jury to decide and the ruling that the conclusion was open to the jury was tantamount to a decision that Google published the paragraphs.
6. There is no case in which it has been held that a search engine operator does not publish such paragraphs after the operator has been notified of them and failed to remove them within a reasonable time. In Metropolitan International Schools Ltd (t/as SkillsTrain and/or Train2Game) v Designtechnica Corp (t/as Digital Trends) Eady J held that Google was not a publisher of such paragraphs before notification or while taking reasonable steps to remove them after notification. Eady J did not go so far as to hold that Google could not be a publisher if it refused to remove them after notification. McCallum J in Bleyer v Google Inc LLC and Fenlon J in Niemela v Malamas each held that Google was not a publisher of such paragraphs before notification but explicitly said that the same conclusion did not necessarily apply after notification. The reasoning of Fenlon J strongly suggests that her Honour would have concluded that Google was a publisher of such paragraphs after notification. In particular, Fenlon J considered that the only real difference between the innocent dissemination test and the passive instrument test involves the burden of proof.[201]

1. Google contends that both principle and authority support the proposition that website forum hosts are not publishers of postings on their websites by third parties even after notification and non-removal, and that the position is similar or a fortiori in respect of paragraphs generated by search engines.
2. Website forum hosts operate websites that invite users to post articles and comments. They are analogous to notice boards in shops where the shop owner invites users to post notices and comments. As a matter of principle, such website forum hosts and shop owners are secondary publishers and liable for defamatory postings if they know of their content and do not remove them. I agree with the analysis of the Hong Kong Court of Final Appeal in Oriental Press Group Ltd v Fevaworks Solutions Ltd in this respect.[202]

310 His Honour also considered the question whether Google republished the content of an external webpage when a searcher clicked on a hyperlink displayed on Google website. He concluded that -

The user does not need to enter the URL into the user's web browser; the Google website is programmed automatically to cause the browser to display the Ripoff Report webpage by clicking on the hyperlink. In these circumstances, Google is a secondary publisher of the Ripoff Report webpage if and to the extent that Google failed to remove the paragraph incorporating the hyperlink after a reasonable time elapsed after effective notification by Dr Duffy.[203]

311 In Duffy, the hyperlinks had additional associated text - not the situation which arose in Crookes. The different conclusions in the two cases are entirely reconcilable.

312 Finally, with respect to the predicted autocompletion which was complained of, Blue J concluded that:

1. The third issue that arises is whether the generation by the Google websites of the words "janice duffy psychic stalker" when a user entered the search term "Janice Duffy" was a publication by Google after notification and failure by Google to remove it within a reasonable time thereafter. On the issue of publication, there is no material difference between the Autocomplete utility and the Related Search utility. The parties do not distinguish between them in their submissions.
2. The words "janice duffy psychic stalker" are generated by Google's programs as a result of Google's programming. The mere fact that the words are programmed to be generated because the user or others have previously searched for those words makes no difference to the physical element. The mental element will be present if Google failed to remove the words within a reasonable time after notification.
3. The only authority on whether a search engine operator publishes words created by the Autocomplete and Related Search utilities is Dr Yeung Sau Shing Albert v Google Inc. In that case, Deputy Judge Marlene Ng held that there was a good arguable case that an operator whose search engine generates objectively defamatory materials by its automated processes is a "publisher" and that Google was more than a passive facilitator vis-à-vis its Autocomplete and Related Search utilities.[204]

313 In the event, his Honour concluded that the defendant was a secondary publisher of the allegedly defamatory material, and that certain defamatory imputations were made out. But he found also that the autocomplete prediction was not defamatory. In the last connection, he said this:

1. [The critical words are] "janice duffy psychic stalker" generated by Google's Autocomplete function.
2. The Autocomplete words are alleged to give rise to the following defamatory imputations:
   1. the plaintiff stalks psychics;
   2. the plaintiff harasses "psychics" by persistently and obsessively pursuing them.
3. The first imputation is not established. The ordinary reasonable person reading the Autocomplete words would understand that they are neither a statement by Google nor a reproduction by Google of a statement by someone else about Dr Duffy. Rather, they comprise a collection of words that have been entered by previous searchers when conducting searches. They do not amount to the proposition that Janice Duffy is a stalker of psychics.

376. The second imputation is not established.[205]

314 Niemela is another interesting decision. The plaintiff relied upon snippets and hyperlinks in connection with an application for a mandatory interlocutory injunction which would compel Google to block 146 URLs containing material defamatory of the plaintiff from appearing as search returns.[206] For its part, Google applied for summary judgment on the basis that it was not a publisher of the snippets and hyperlinks, or alternatively to strike out parts of the claim. This was a pre-notification case.

315 Madam Justice Fenlon refused the application for an injunction without the need to decide the publication point.[207] Her Honour acceded to Google's summary judgment application. The plaintiff relied upon four causes of action, but it is only necessary to mention what her Honour held with respect to the claim in libel.

316 The material facts of the matter, not in dispute, were these:

...

(a) Search results and "snippets" on Google's websites are generated automatically through the operation of computer algorithms in response to search terms inputted by users.

(b) Google's proprietary algorithm is programmed by Google to rank search results according to their probable perceived relevance to users.

(c) Google maintains different search platforms for different countries and search results may vary from platform to platform.

(d) The search results generated by the algorithm are generated from the automated review of more than 60 trillion websites. They are continuously updated and may vary from hour to hour or even from minute to minute.

(e) Google's search platforms provide a means for internet users to locate websites hosted by third parties that may be of interest to the user.

(f) Google does not promote or endorse particular search results. It neither warrants the reliability of websites generated in search results nor cautions the user that the authors of statements found on websites may not be trustworthy.

(g) Google does not amend search results for commercial gain.

(h) A single page of search results generally displays 10 results, with hyperlinks to third party websites accompanied by snippets of text from those sites. More results are displayed on further pages.

(i) Pages may include third party advertising which is identified as such.

(j) Search results reflect the content of third party websites at the time the sites were last crawled by the computers processing Google's search algorithm. Changes in the site by the third party host may not be reflected in search results until the page is crawled again by Google's computers processing the algorithm. Google does not control the content of third party websites, nor changes to those websites.[208]

317 As to publication, her Honour first dealt with the hyperlinks relied upon by the plaintiff. In reliance upon Crookes, she held that the material in the hyperlinked webpages was not thereby published.

318 With respect to the snippets, and having noted the contrary conclusions expressed in Trkulja (No 5) and Metropolitan Schools, her Honour analysed the position this way:

1. As I have noted in disposing of Mr. Niemela's claim that Google is a publisher of the hyperlinks contained in its search results, Metropolitan was considered by the Supreme Court of Canada in Crookes, a defamation case originating in British Columbia. At para 21 Abella J speaking for six judges approved the approach taken in Bunt v Tilley and Metropolitan:

[21] Recently, jurisprudence has emerged suggesting that some acts are so passive that they should not be held to be publication. In Bunt v Tilley, [2006] EWHC 407, [2006] 3 All ER 336 (QB), considering the potential liability of an Internet service provider, the court held that in order to hold someone liable as a publisher, "[i]t is not enough that a person merely plays a passive instrumental role in the process"; there must be "knowing involvement in the process of publication of the relevant words" (para 23 (emphasis in original); see also Metropolitan International Schools Ltd v Designtechnica Corpn, [2009] EWHC 1765, [2011] 1 WLR 1743 (QB)).

1. In Crookes there were two additional judgments concurring in the result but differing slightly in their analysis. Yet all of the judges in Crookes accepted the principle that publication requires a deliberate act in relation to the words in question. I have already referred to the reasons of Abella J on this point at para 26 of Crookes.
2. In concurring reasons, at paras 50 and 51, McLachlin CJ and Fish J advanced an alternative publication standard for hyperlinks:

[50] In sum, in our view, a hyperlink should constitute publication if, read contextually, the text that includes the hyperlink constitutes adoption or endorsement of the specific content it links to.

[51] It is true that the traditional publication rule does not require the publisher to approve of the material published; he or she must merely communicate that material to a third party. However, the proposed adoption or endorsement standard for references is conceptually different. A mere reference without any adoption or endorsement remains that - a content-neutral reference.

1. Deschamps J, in her concurring reasons, also noted that proof of publication is necessary in order to establish liability for defamation. At para 55 she referred to the traditional test for publication:

[P]ublication has two components: (1) an act that makes the defamatory information available to a third party in a comprehensible form, and (2) the receipt of the information by a third party in such a way that it is understood.

1. At para 56 she framed the issue before the court as "whether the first component of publication needs to be reconsidered owing to the impact of new forms of communications media."
2. Deschamps J concluded that only deliberate acts can lead to liability:

[59] A more nuanced approach to revising the publication rule, and one that can be applied effectively to new media, would be for the Court to hold that in Canadian law, a reference to defamatory content can satisfy the requirements of the first component of publication if it makes the defamatory information readily available to a third party in a comprehensible form. In addition, the Court should make it clear that not every act, but only deliberate acts, can lead to liability for defamation. [Emphasis in original.]

1. In considering these passages in Crookes I am mindful that the narrow ratio of that case is that providing a hyperlink does not constitute publication of the article to which the hyperlink leads. Crookes did not address snippets and indeed Abella J stated at para 42:

[42] Making reference to the existence and/or location of content by hyperlink or otherwise, without more, is not publication of that content. Only when a hyperlinker presents content from the hyperlinked material in a way that actually repeats the defamatory content, should that content be considered to be "published" by the hyperlinker. [Emphasis added.]

1. A snippet repeats words from the hyperlinked article and, as in the present case, those words can be defamatory. For example, the following text is taken from one snippet displayed in the Google search results of Mr Niemela's name: "Glenn J. Niemela complaint review: Glenn J. Niemela he's a scam artist and steals from his clients while doing nothing to help them with their ...".
2. Using the traditional approach to publication, Google's lack of awareness of the content of the words it repeated on a search page snippet would not prevent it from being a publisher, although it would likely be able to avail itself of the defence of innocent dissemination after the prima facie test for defamation had been made out. As Abella J noted at para 18 of Crookes:

[18] Under this sole disseminator/sole reader paradigm, the breadth of activity captured by the traditional publication rule is vast. In R v Clerk [1744] EngR 1309; (1728), 1 Barn KB 304, 94 ER 207, for example, a printer's servant, whose only role in an act of publication was to "clap down" the printing press, was found responsible for the libels contained in that publication, despite the fact that he was not aware of the contents (p 207). In Hird v Wood (1894), 38 SJ 234 (CA), pointing at a sign displaying defamatory words was held to be evidence of publication. Other cases have also held that acts merely facilitating communication can amount to publication: see, eg, Buchanan v Jennings, [2004] UKPC 36, [2005] 1 AC 115; Polson v Davis, 635 F Supp 1130 (D Kan 1986), aff'd [1990] USCA10 53; 895 F 2d 705 (10th Cir 1990); Crain v Lightner, 364 SE 2d 778 (W Va 1987), at p 785; and Spike v Golding (1895), 27 NSR 370 (SC in banco). And in McNichol v Grandy, the defendant was found to be liable when he raised his voice and made defamatory statements that were overheard by someone in another room.

1. As I have noted, however, the tenor of Crookes and of recent jurisprudence in England is to narrow the test for who is a publisher of defamatory material to those who do deliberate acts. In Canada this shift originates in the Supreme Court of Canada's recognition post-Charter "that what is at stake in an action for defamation is not only an individual's interest in protecting his or her reputation, but also the public's interest in protecting freedom of expression": Crookes at para 31, citing Hill v Scientology.
2. It is worth noting at this point that the English approach requiring a deliberate act to establish publication (the passive instrument test) is related to, but distinct from, the defence of innocent dissemination which is a well-settled principle in Canadian defamation law.
3. As described in Crookes at para 20 and in Society of Composers, Authors and Music Publishers of Canada v Canadian Assn of Internet Providers, 2004 SCC 45 at para 89, a defendant in a subordinate publisher role can avail itself of the defence of innocent dissemination by showing that it:
   1. had no actual knowledge of an alleged libel;
   2. is aware of no circumstances to put it on notice to suspect a libel; and
   3. committed no negligence in failing to find out about the libel.
4. In short, the defence is available to defendants who, while not the creator of the defamatory material, disseminate the material in a manner that would ordinarily amount to publication save for the innocent nature of the dissemination.
5. In substance, the passive instrument test appears to have requirements very similar to the defence of innocent dissemination. However, it shifts the burden of proof from the defendant to the plaintiff. Where innocent dissemination is pleaded, the defendant bears the burden of showing that it had no actual knowledge of an alleged libel, was aware of no circumstances to put it on notice to suspect a libel, and committed no negligence in failing to find out about the libel.
6. On the other hand, if the passive instrument test is applied the plaintiff bears the burden of showing that the defendant was not a mere passive instrument because at the initial prima facie stage the plaintiff must prove the third element of publication.
7. The majority in Crookes did not expressly adopt the passive instrument test for publication, but Abella J endorses Metropolitan which applied it. Furthermore, much of the majority's reasoning for concluding on public policy grounds that hyperlinks should not be treated as publication, applies with equal or greater force to search results. Abella J observed at paras 34 and 36:

[34] The Internet's capacity to disseminate information has been described by this Court as "one of the great innovations of the information age" whose "use should be facilitated rather than discouraged" (SOCAN, at para 40, per Binnie J). Hyperlinks, in particular, are an indispensable part of its operation. As Matthew Collins explains, at para 5.42:

Hyperlinks are the synapses connecting different parts of the world wide web. Without hyperlinks, the web would be like a library without a catalogue: full of information, but with no sure means of finding it.

(See also Lindsay, at pp 78-79; Mark Sableman, "Link Law Revisited: Internet Linking Law at Five Years" (2001), 16 Berkeley Tech LJ 1273, at p. 1276.)

...

[36] The Internet cannot, in short, provide access to information without hyperlinks. Limiting their usefulness by subjecting them to the traditional publication rule would have the effect of seriously restricting the flow of information and, as a result, freedom of expression. The potential "chill" in how the Internet functions could be devastating, since primary article authors would unlikely want to risk liability for linking to another article over whose changeable content they have no control. Given the core significance of the role of hyperlinking to the Internet, we risk impairing its whole functioning. Strict application of the publication rule in these circumstances would be like trying to fit a square archaic peg into the hexagonal hole of modernity.

1. There are hundreds of millions of active websites over the Internet and trillions of webpages. Search engines make the Internet a viable and effective information and communication resource. The Internet cannot be successfully navigated without search services such as those Google provides. If hyperlinks are the pathways search engines are the maps. Without snippets, which give a sample of the text in which the search terms are imbedded, navigating the Internet would be much more difficult. Users would have to click on each URL, access the hyperlinked webpage, and examine its contents to determine relevance.
2. Adoption of the passive instrument test for publication in British Columbia was addressed by Burke J in Weaver v Corcoran. The plaintiff in that case sued over reader comments posted about him on a forum hosted by the National Post. After referring to Bunt v Tilley and Metropolitan as well as Tamiz v Google Inc, Burke J concluded at para 282 that the jurisprudence establishes that "some awareness of the nature of the reader post is necessary to meet the test of publication." She continued at para 284:

Until awareness occurs, whether by internal review or specific complaints that are brought to the attention of the National Post or its columnists, the National Post can be considered to be in a passive instrumental role in the dissemination of the reader postings. It has taken no deliberate action amounting to approval or adoption of the contents of the reader posts. Once the offensive comments were brought to the attention of the defendants, however, if immediate action is not taken to deal with these comments, the defendants would be considered publishers as at that date. [Emphasis added.]

1. In Weaver Burke J found that it was not reasonable to expect the defendant to pre-vet thousands of visitors' comments for defamatory content and that the defendant did not therefore have the awareness of the words necessary to meet the test of deliberate publication: at paras 273, 282. That reasoning applies all the more to a search engine that automatically trawls trillions of webpages in seconds.
2. Added to the sheer volume of material is the obvious difficulty of developing programming to detect and screen out defamatory words. It is apparent that a search engine could not simply be programmed to block every site containing, for example, the words "scam artist" or "steals" without blocking millions of pages of non-defamatory content.
3. Google programs its search algorithm so that it locates URLs likely to relate to a user's search query. It is not aware of the snippets and hyperlinks produced, nor can it be, realistically. In the words of Eady J in Metropolitan, Google does not authorize the appearance of the snippets on the user's screen "in any meaningful sense" but "has merely, by the provision of its search service, played the role of a facilitator": at para 51.
4. In summary on this issue, I conclude that Google is a passive instrument and not a publisher of snippets. There is accordingly no issue for trial in relation to defamation.
5. I emphasize that I have not been asked in this case to consider whether Google could be a publisher of snippets and search results after notice of defamatory content.[209]

319 In all, in our view, the judicially-expressed position with respect to publication by search engines can be summarised this way:

(1) Only two possibilities have been canvassed. First, that a search engine is simply not a publisher of search results, hyperlinked external webpages, or autocomplete predictions. Second, that a search engine is a secondary publisher of such matter,[210] almost certainly protected by the innocent dissemination doctrine until notice is given, and for a reasonable time thereafter. Then, if it does not 'take down' the impugned matter, it will lose the innocent dissemination protection.

(2) With the possible exception of the decision from which this application is brought, no authority to which we were referred suggests that a search engine is to be regarded as a primary publisher from the moment that search results are displayed. There has been no resort to a Thompson analysis.

(3) The two possible approaches are not related to the Byrne line of authority, by which, in certain circumstances, a person may become a primary publisher.[211]

(4) The logical consequence of the first approach is that a search engine would not be a publisher, whether before or after notice. That proposition is inherent, though not plainly stated, in the judgment of Eady J in Metropolitan Schools.

(5) In the two instances where the plaintiff sued upon autocomplete predictions, the decisions[212] did not distinguish between the material thus displayed and the display of search results from external websites. There are differences between the two types of material displayed. They were not explored in this Court. No attempt was made by either party to contend that the publication point should be differently resolved in the two situations.

(6) In a number of contexts to do with search engines, it has been stated or hypothesised that the position of a search engine may alter if notice of alleged defamatory matter is given and no action is taken within a reasonable time. In Metropolitan Schools, Eady J recognised the difference between the positions of a forum provider and a search engine proprietor. Nonetheless, he devoted considerable attention to Google's 'take down' policy, and its implementation. But what seems not to have been analysed is how the conception of 'take down' in response to a notice works in the case of a search engine.

320 A plaintiff sues on allegedly defamatory matter, published on one or more defined occasions. The matter must be particularised. In the simple case, the defendant may remove the defamatory notice or poster (and so avoid becoming a primary publisher in the Byrne and Urbanchich kind of case), or may cease to sell a newspaper edition or lend a book (the innocent disseminator line of cases). Then, in case of posts on a web forum, again there is fixed and continuing matter, present until removal. But when a plaintiff sues, as here, solely upon the results produced by a search engine - and does not seek to rely upon the content of hyperlinked webpages - it is those results, produced at a moment in time, which are the allegedly defamatory matter. Those results cannot be 'taken down' - whether they be images, snippets, URLs or autocomplete predictions. The most that could be done is to prevent a possible re-published defamation in the future - that is, by the return of results in response to a new search enquiry. If it be assumed that search results are a secondary publication, in our opinion it is unclear how an innocent dissemination defence could be lost in consequence of the giving of notice in the circumstances here under discussion. If, contrary to the authorities thus far, a search engine was to be treated as having the potential to become a primary publisher after notice, the same problem would arise.

321 The conclusions just expressed, to be clear, do not address the case in which a plaintiff relies upon the content of a hyperlinked webpage. But different problems might arise in that context. If the content of an external webpage was downloaded by the searcher clicking on a hyperlink, a plaintiff might seek to rely upon it. But because the content of the world wide web is subject to constant change, it is not clear that giving notice of allegedly defamatory material on a third party website would address the defamation complained of. Neither does it seem clear that 'taking down' a hyperlink would bear upon the search engine's possible liability for publication, 'post-notice', of the defamatory matter which is complained of.

322 Despite our doubts, however, we think that in deciding the present application we should take it to be arguable that notice could be a consideration relevant to publication (whether as a primary or secondary publisher) of matter of the kind upon which the plaintiff relies.

323 That takes us to the contest between the two approaches to which we referred at [319](1) above.

324 The physical element of publication is satisfied when a search engine displays web search results. That is not less the case because a search engine responds to a user's inputting of search terms, and does so by producing material drawn from external websites. The search engine itself disseminates information.

325 It is not in question that there is a mental element to publication - that is, that the defendant has an intention to publish the material complained of to a person other than the parties.[213]

326 Historically, the innocent publisher - newspaper vendor, library or the like - has been protected by the innocent dissemination doctrine. The analysis is that such a person, standing in a distribution chain, physically publishes and intends to publish to third parties the newspaper or book within which the defamatory matter is to be found, but in doing so does not know and is not careless of the fact that the defamatory matter is present.

327 In the situations just mentioned, the innocent publisher does nothing at all to gather together, or reassemble, the defamatory matter.

328 At a factual remove from the situations just mentioned is the telephone company or postal service[214] which may carry, in the former case, oral defamatory material, and in the latter case, a document containing defamatory material. Statute aside, such entities do no more than intend that whatever it may be that A wishes to say or send to B is conveyed. Yet, according to Gatley on Libel and Slander, at common law they might be accounted secondary publishers, though in all probability, innocent disseminators.[215] This gives life to the concept of relevant intention.

329 When one comes to the question of publication on the internet, the starting point, for present purposes, is Bunt. That is where the idea of an entity 'performing no more than a passive role in facilitating postings on the internet' and for that reason not being a publisher, arose. It is to be noted that Eady J would put a telephone company in the same position. Passivity, on his Lordship's analysis, contraindicated 'knowing involvement in the process of publication of the defamatory matter', or 'awareness or at least an assumption of general responsibility'.

330 As we have previously observed, it would potentially be misleading to treat 'publication on the internet' as if all situations were one and the same. Whilst the situation in Bunt was as close to the telephone carrier or postal service as one could get in an internet context, a different problem arises where the defendant's conduct goes beyond merely providing access to the internet. In Tamiz, at first instance, Eady J treated Google's role, as provider of the platform blogger.com, as 'a purely passive one'; and so Google was not to be regarded 'as a publisher or even as one who authorises publication, under the established principles of the common law'.[216] But the Court of Appeal, whilst not holding that there was any fundamental error in the Bunt analysis, concluded that Eady J had been wrong to regard the defendant's role in the instant case as a purely passive one.[217] Google facilitated the blogs. Nonetheless, according to the court, Google was not the primary publisher, and very doubtfully a secondary publisher. But, after notification, the principles in Byrne applied.

331 The alternative analysis, with respect to forum hosts, preferred in Fevaworks, is that the defendant was a secondary publisher from the date of a post onwards, but protected in accordance with the innocent disseminator doctrine.

332 The search engine variant involves different considerations again. Unlike an ISP 'conduit' or the providers in Tamiz and Fevaworks, a search engine takes an active role in producing search results. Its role, in our opinion, could not be characterised as a passive one. We respectfully doubt that its role can be described as being simply a 'facilitator', in the sense of merely enabling contact between A and B. There seems to us to be less reason than in either the Bunt or Tamiz situations to hold that a search engine is not a publisher of search results (again, we confine ourselves to results such as the plaintiff relies upon here).

333 The search for consistency in principle, referred to by Eady J in Tamiz, cannot deflect attention from the particular features of a search engine. The Court of Appeal decision in Tamiz, tailored to the features of that case, is an instance of principle trumping a 'one size fits all' search for consistency.

334 The decision in Metropolitan Schools is the juridical rationale for the 'no publication of search results' conclusion. Essentially, Eady J considered two internet cases - Godfrey and Bunt. Neither of them was a search engine case. Each of them involved an ISP. Relevantly, the ISPs undertook different functions.

335 Justice Eady recognised that the earlier cases did not decide the question for his determination. At the heart of his judgment was the proposition that:

Of course, to be liable for a defamatory publication it is not always necessary to be aware of the defamatory content, still less of its legal significance. Editors and publishers are often fixed with responsibility notwithstanding such lack of knowledge. On the other hand, for a person to be held responsible there must be knowing involvement in the process of publication of the relevant words. It is not enough that a person merely plays a passive instrumental role in the process. (See also in this context Emmens v Pottle (1885) 16 QBD 354 at 357 per Lord Esher MR.)[218]

His Lordship then cited a passage from the judgment of Lord Esher MR in Emmens - which was, of course, the starting point for the doctrine of secondary publication and innocent dissemination.

336 This led on to His Lordship's observation that -

When a snippet is thrown up on the user's screen in response to his search, it points him in the direction of an entry somewhere on the web that corresponds, to a greater or lesser extent, to the search terms he has typed in. It is for him to access or not, as he chooses. It is fundamentally important to have in mind that the third defendant has no role to play in formulating the search terms. Accordingly, it could not prevent the snippet appearing in response to the user's request unless it has taken some positive step in advance. There being no input from the third defendant, therefore, on the scenario I have so far posited, it cannot be characterised as a publisher at common law. It has not authorised or caused the snippet to appear on the user's screen in any meaningful sense. It has merely, by the provision of its search service, played the role of a facilitator.[219]

and to his likening what was done to the use of a library catalogue. But whereas a catalogue which records a defamatory statement in a book might give rise to liability under the 'repetition' rule, yet the repetition of defamatory allegations in a snippet, his Lordship concluded, involved no conscious choice by the search engine; and that was important.

337 Although His Lordship referred to the Google search engine playing 'the role of facilitator', it appears to us that his decision differed from his analysis in Bunt and in Tamiz. It rested on the idea that the mental element of intent to publish was not present because the search results were the automated consequence of external input of a search term and the automated response by reference to external webpages.

338 We pause to make this clear. The decision in Bunt could be accepted and the decision in Metropolitan Schools rejected, just as the Court of Appeal's decision in Tamiz could be accepted and the decision in Metropolitan Schools rejected.

339 The other decisions which have held 'no publication' in a search engine case can be shortly mentioned. We have dealt with them at length earlier in these reasons. In Bleyer, McCallum J accepted 'the English line of authority' that, at least before notification of complaint, Google could not be liable as a publisher of search results. A difficulty is that 'the English line of authority' was a collation of different circumstances and somewhat different reasoning. In Niemela, Fenlon J separated out the hyperlinks and the search results for distinct consideration. Her Honour appears to have regarded that as having been necessary in light of Crookes. We doubt that this was so. But be that as may, her Honour recognised that -

Using the traditional approach to publication, Google's lack of awareness of the content of the words it repeated on a search page snippet would not prevent it from being a publisher, although it would likely be able to avail itself of the defence of innocent dissemination after the prima facie test for defamation had been made out.[220]

340 But, her Honour concluded, the tenor of Crookes and 'recent jurisprudence in England' had been to narrow the test of who is a publisher 'to those who do deliberate acts' - a shift in Canada which reflected the Canadian Charter of Rights and Freedoms.[221]

341 Further, her Honour pointed out the utility of search engines, and the negative consequences of stifling their use.

342 Her Honour's conclusion that the Google search engine was a 'passive instrument' was not, we think, exactly the approach of Eady J in Metropolitan Schools, although his Lordship certainly referred to Google playing 'the role of a facilitator'. Moreover, as we have said earlier, we do not think it is correct to describe the role of a search engine as a 'passive instrument'.

343 In Trkulja (No 5), Beach J departed from the reasoning in Metropolitan Schools on the basis that intention to publish might have been found by the jury on the footing that the search engine produced the results which it was intended - by programming - to produce. He likened that to the position of a newspaper seller or librarian.

344 It is not clear to us that the newspaper seller or librarian analogy is apt. The seller distributes that with which he or she has been provided - we assume unread - in the form in which it was provided. The librarian lends out a book - we assume unread - which has come to the library in completed form. The search engine, however, receives a request for information and, by operation of an automated process, responds by producing from external webpages a very short note of what it perceives may be useful to the searcher. This involves something different to a librarian being asked where to find a book about a particular subject, and directing the enquirer to a particular shelf. Whilst the search results, by analogy, say something about what is in a book or books on a shelf, because the user's search term is understood by the search engine as a string of words, the response may direct the user, to continue the analogy, to an irrelevant book on a shelf.

345 The ruling of Beach J has been acted upon in Yeung and Duffy. It has attracted varying academic comment. Riordan, for instance, says that -

Conversely, in Trkulja v Google Inc LLC (No 5) the Supreme Court of Victoria concluded that it was open to a jury to conclude that Google was a publisher and stood in an analogous position to a newsagent who sold a newspaper containing a defamatory article. To the extent that Trkulja imposed prima facie liability upon Google for mere operation of its indexing algorithm, and before it had actual knowledge of the defamatory publication, Trkulja is - with respect - inconsistent with persuasive authority and clearly wrong. In Duffy v Google Inc, the Supreme Court of South Australia followed Trkulja and concluded that Google was a publisher of snippets, at least where it failed to remove them within a reasonable period of notification. In that case, Blue J treated Google as being in the same position as a host of defamatory material; this approach is open to question, since it ignores important differences between these network and application layer intermediaries.[222]

346 Whilst we are not persuaded of the correctness of the entirety of the reasoning of Beach J, it does not follow that his Honour's overall conclusion is impugned. For, as we have said, we are not persuaded that a search engine's role is to be characterised as 'passive' and as being no more than one of 'facilitation'.

347 In our view, for reasons which we have attempted to explain, no analogy is entirely apt, and no one internet decision logically leads to another. We consider that the sound approach is to approach the matter by reference to first principles, as expounded in Webb. That was the preferred approach of Blue J in Duffy and first principles were the starting point for the analysis by Fenlon J in Niemela. Her Honour was led away from that analysis by what she understood to be the import of recent jurisprudence, and by concerns about the chilling effect of holding that search engines, which have very great utility in the age of the internet, are publishers of search results.

348 On first principles, we consider that a search engine, when it publishes search results in response to a user's enquiry, should be accounted a publisher of those results - and in this we include autocomplete predictions. It is a participant in a chain of distribution of material.

349 We also consider that a search engine should be accounted a secondary publisher. It is true that its automated response picks up words used in the search term and in identified webpages. But that adds nothing to what has already been published. So far as the repetition rule has any part to play, it does not lead to a conclusion that the search engine is a primary publisher of content produced by a word search in response to a searcher's inputted query. Google's knowledge of that content is essentially confined to an understanding that the content will have some connection with the inputted search terms. It lacks the characteristics of a primary publisher which Ribeiro PJ described in Fevaworks.[223]

350 The argument in this Court and below understandably focused upon intention to publish, that being characterised as the point of departure between Metropolitan Schools and Trkulja (No 5). But we think that too much has been made of the question of intention, narrowly described, as the determinant whether Google publishes, in the legal sense, that which it in fact publishes to third parties.

351 We live in a world where automated action is becoming, if it is not already, the norm. From automated processes in a factory to the pervasive 'robo' calls during election campaigns, computer-generated programs are at work. Yet no one would think of saying that a car manufacturer does not intend to build a car, in the way in which it is constructed, because steps in the process are automated, performed in accordance with a computer program. And take the librarian situation. The place of a book on a library shelf, in accordance with its Dewey classification, has characteristically involved human assessment of that book's content. But today, very many books can be accessed online. What if a librarian used a computer-designed automated system to give a book its Dewey classification? Would the file card, produced in such a way, not be published by the library because of the absence of direct human intervention? Would the librarian no longer (probably) be a secondary publisher? So to conclude would be, we think, a strange outcome. Would it be different if the computer program was not directly the work of human composition, but was designed by a computer[224] - the element of human intervention being a further step removed?

352 We appreciate that the examples of computer-mediated actions which we have just mentioned do not precisely equate with the work of the Google search engine. For in each of the postulated situations the program is not, or would not, be designed to produce a particular result in response to an externally made request by resort to externally created material. But the examples do suggest that the fact that the defamatory matter complained of is the product of an automated response does not necessarily gainsay an intention to publish that material. When that consideration is supplemented by the facts that the Google search engine holds itself out as providing a means of navigating the web, that its role, in our judgment, is not passive and that in providing a search result it does more than merely facilitating contact between A and B, it seems to us that intention to publish that which is in fact published is an available conclusion. Concerns that thereby a search engine may become liable in defamation proceedings when that might be regarded as an unsatisfactory consequence are met, in our view, if it is concluded, consistently with those authorities which have considered the issue, that a search engine is a secondary publisher.

353 Assuming that it is correct to say that a search engine is a secondary publisher of search results of the kind in issue in this matter, we think that an innocent dissemination defence will almost always, if not always, be maintainable in a period before notification of an alleged defamation.[225] Further, for reasons which we have attempted to explain, we have reservations as to whether, and how, notification of a past defamatory publication by way of search results could lead to innocent dissemination becoming something else. Nonetheless, for reasons which we have explained, we treat it as being arguable that notification can have some part to play upon the question of innocent dissemination.

354 Having regard to the way in which the plaintiff advanced his case, and perhaps the way in which the judge disposed of the matter below, it is necessary to say something about the possibility of the Google search engine being a primary publisher of search results, either from the outset, or perhaps on a Byrne basis. As we have said, no authority, either directly or contingently, supports such an approach.

355 For our part, having regard to the circumstances which we have described in detail, neither the approach of the plurality nor of Gaudron J in Thompson could lead to the search engine being held a primary publisher from the outset. Ability to control and supervise the word search results gleaned from more than 63 trillion individual webpages in a fraction of a second would seem out of the question, and the concept of authorisation seems alien to what occurs.

356 What, then, of Byrne and Urbanchich? The only way that the particular line of authority could be in point is if the search results which are displayed could be treated as having been displayed unbeknownst to Google - like the notice on the golf club wall. The difficulty with that is that it is Google itself which displays the results. But if that hurdle could somehow be overcome, there would be the problem about notice to which we have already referred.

357 In all, in our opinion, the secondary publisher/innocent dissemination defence analysis appears to be both the preferable outcome in point of principle, and to be a rational way of dealing with the problem of results produced by a search engine. It follows that it should be accepted that the production of results by an internet search engine does not constitute a primary publication of those results. The operator of the search engine may be liable, if at all, for any defamatory content in the search results (in which we include autocomplete predictions) only as a secondary publisher.

358 That takes us to resolution of the question of publication in the present case. In more than one respect, the situation is unsatisfactory.

359 First, the argument below was entirely conducted by counsel for Google. That was not Google's fault, but it is the fact. The judge quizzed counsel at length, but his Honour did not have the advantage of contrary argument.

360 Second, it is the fact, and again it is no fault of counsel for Google, that argument below was conducted on the footing that the plaintiff pleaded that Google was a primary publisher. Notice was referred to by counsel in argument, but not in the context of secondary publication and the innocent dissemination defence. The structure of the ASoC conduced to such an approach.

361 Third, the precise basis upon which the judge decided Google's application is not altogether clear.

362 On the one hand, his Honour stated that he acted on the ruling of Beach J in Trkulja (No 5). As we have attempted to make clear, that case was founded on the proposition that Google, if a publisher at all, was a secondary publisher, to whom an innocent dissemination defence was available.

363 On the other hand, although his Honour noted Google's acceptance that it bore the onus of establishing that 'in no circumstance, whether on notice or not, can Google be a publisher', the reasons are otherwise silent upon the question of notice. And yet, if Google was regarded as a secondary publisher, notice would be of very considerable significance.

364 In the present case, as we have already noted, a number of the pages particularised with respect to both the images matter and web matter predated the notice of 3 December 2012, this bearing upon the likely ambit of any maintainable claim if Google was a secondary publisher. It might be responded, however, that his Honour did not need to deal with notice and its consequences once he had concluded that Google was a publisher of the impugned matter, even assuming that he concluded that it was a secondary publisher.

365 But against such a reading of his Honour's reasons are - (1) his statement that 'Mr Trkulja does not allege that Google is a subsidiary or secondary publisher of defamatory material which has been published by a third party',[226] (2) his reference to what Thompson established respecting primary publication,[227] and (3) his statement that 'Applying the reasoning of the High Court in Thompson, Google's ignorance of the specific material generated in response to any given user in query does not prevent it from being found liable as a publisher'.[228]

366 In all, we think it more likely than not that his Honour found that it was arguable that Google was a primary publisher, not on a Byrne/Urbanchich basis, but from the moment when the search results and autocomplete predictions respectively were displayed on the searcher's screen.[229]

367 Fourth, we refer again to what occurred in this Court. Senior Counsel for the plaintiff made it clear that it was his client's case that Google was a primary publisher. Faced with Google's objection to an amendment which would have squarely raised a case based on secondary publication and loss of the innocent dissemination defence by reason of notice, counsel simply did not pursue the proposed amendment. He could have done so, had that been the plaintiff's intended argument.[230]

368 It appears to us to flow from the circumstances which we have described that we are confronted by a situation in which the ASoC does not squarely allege facts pertinent to a secondary publication case against Google; where counsel abandoned a proposed pleading amendment which would have achieved that result; and where counsel affirmed that it was the plaintiff's case that Google should be held liable as a primary publisher of the impugned matter.

369 We have used the word 'unsatisfactory' a number of times in this part of our reasons, and it can sensibly be used in another connection also. It would be unsatisfactory, from Google's standpoint, if its application failed because the plaintiff might have advanced a different case in opposition to that application.

370 Having concluded that the basis upon which the plaintiff seeks to pursue his case that Google published defamatory matter is not arguable,[231] it appears to us that service might be set aside on that basis alone. That also would be unsatisfactory, because we do consider, had the plaintiff articulated his case on the basis of secondary publication, that such basis would have been fairly arguable.

371 It might be, if service was set aside, that the plaintiff would be able to re-plead, re-cast his argument, re-serve Google out of the jurisdiction, and then await the inevitable challenge to such service. Alternatively, absent any other problem, it might be that the Court would give the plaintiff the opportunity of re-pleading, and of advancing a different case.

372 Ultimately, however, the difficulties of which we have been speaking are not a live issue. That is because, in our opinion, the allegedly defamatory matter upon which the plaintiff relies is incapable of conveying a defamatory meaning. It is to that issue which we now turn. We pause only to note that the possible issue estoppel to which reference was made by the judge would face considerable difficulty of application on the case advanced for the plaintiff, whatever other difficulties it might have. That is because, again, the issues in Trkulja v Google (No 5) concerned secondary publication whereas the plaintiff now pursues a case based on primary publication.

Grounds 9-10. Is the matter complained of capable of defamatory meaning?

373 We have already described, in some detail, the content of the images matter and web matter upon which the plaintiff sues.[232] We have referred also to the inputted search terms,[233] and to the question whether the alleged publication included the inputted search terms.[234] Again, we have referred to the defamatory imputations allegedly conveyed by the images matter and web matter as false innuendo,[235] and to a pleading of true innuendo with respect to the images matter.[236] The question is whether, as pleaded, the images matter or web matter is capable of conveying a defamatory meaning.

374 The judge's resolution of that question was succinct. Earlier, we have set out the relevant part of his Honour's reasons.[237] As we have noted, his Honour's conclusion fastened upon the compilation of four images which is a part of page four of Annexure B to the ASoC.

375 It was submitted for Google in writing that the 'ordinary reasonable individual user' of the search engine would know that the indicia received in response to a query are 'no more than indicators of what some of the materials that respond to the query might be, and where those materials can be located on the web'. The ordinary reasonable user 'would not conclude that those abstract indicia are intended to be understood as true'. Further, in Duffy, Blue J had held that autocomplete predictions were not capable of conveying a defamatory meaning to the ordinary reasonable user of the search engine; and like reasoning should apply to compilations of images.

376 Orally, counsel submitted that the level of understanding of the ordinary reasonable user of the search engine was

at least the level attributed by the High Court in [ACCC], which is an understanding that the search results are the product of a search engine which is entirely automated. The results are the product of the user's search terms, and the results are in the nature of references only, references to material that is found elsewhere in the web, that is decontextualized.

377 There was this interchange between counsel and the Bench:

COUNSEL: ... so the ordinary reasonable person would suspend judgment about inferences or assumptions about the connection between one image and another because they have the immediate ability to get more context concerning that matter. So why would they jump to judgment? That would not be a reasonable viewer's stance about the matter.

ASHLEY JA: And the same with autocompletions?

COUNSEL: Yes, your Honour, the same. An autocompletion a reasonable person would not consider is a statement that they would attribute any defamatory meaning or inference to. I guess all of this emphasises, your Honour, that context is singularly important in this question of capacity to convey a defamatory meaning, and there is a real question about how much context and how deep. But stopping at the level that the High Court thought was significant, for its purposes, in the misleading case, that is sufficient for our purposes.

We know what the High Court attributed to an ordinary user of the internet in the Google case. It is supported by the material here and, indeed, it is a fortiori when you have got images, you have simply got a collection of photographs, and a person receiving that result will know that you can't jump to any conclusion about the connection between one photograph and the next. Some might recognise the former Chief Commissioner of Police, some might recognise Jill Meagher, et cetera, et cetera. No ordinary person would jump to the defamatory imputation or inference that is alleged.

...

Our submission is an ordinary reasonable user of a search engine - who it must be remembered has entered the search query in the first place, will have the degree of knowledge of the operation of the internet and the nature of search results, and the ability to contextualise them by clicking on them - that person would not think less of anyone because of the automated display by the search engine of a collection of images, or the automated completion of a search query that is partly completed.

Now, if we are right about that, we would submit that the proper conclusion to reach is that the collective or composite images relied upon are incapable of supporting a finding of defamation and, therefore, on that ground the case should be stopped.

378 It was submitted in writing for the plaintiff that 'capability' of conveying a defamatory meaning was not raised below. It had only been submitted that the alleged publication 'was not defamatory'.

379 It was next submitted that it mattered not that the words were written by a human or by operation of an algorithm. The important question is what the words which are complained of mean.

380 It was further submitted that judicial notice could not be taken, and there was no proper evidence, of the extent of knowledge of the nature of search engine results which, as Google would have it, were to be attributed to the ordinary reasonable user of the search engine.

381 Orally, counsel submitted that the 'capability' argument advanced for Google was not the 'traditional ... submission advanced in cases such as Favell'.[238] It was a 'rather contrary discrete submission'. Further, there was a question whether 'the Court should, in any way, place significance on the way in which the publication reaches the third party'. That was not something to which the common law had given any significance in the past.

382 Upon the question whether Annexures A and B to the ASoC - the particularisation of the defamatory matter pleaded by paragraphs 16 and 18 of the ASoC - were to be read as a whole, counsel submitted that the jury would be entitled to look at any one of the particularised pages and conclude that it was defamatory of the plaintiff. So, respecting the images matter, the jury was not bound to look at all 20 pages and try to draw an imputation from it. Counsel referred to Burrows v Knightley.[239] Counsel submitted also that the defendant would not be entitled to point to other pages in either Annexure to show that images displayed were not necessarily of Melbourne criminals at all.

383 Counsel for Google submitted that the question whether words complained of have a capacity to defame is a question of law for the Court, not a jury question. He further submitted that the publication must be taken as a whole. Clearly, a composite publication was relied on. The question was what single meaning the entirety of the publication conveyed. There is a limited ability to dissect a matter artificially in pleadings.

384 In Favell, Gleeson CJ and McHugh, Gummow and Heydon JJ said this:[240]

Bearing in mind the dual nature of the application to Helman J, seeking both a striking out of certain paragraphs in the amended statement of claim, and the entry of summary judgment for the respondents, on the ground that the pleading disclosed no reasonable cause of action, the question for decision was whether the material published was capable of giving rise to the defamatory imputations alleged. In the Court of Appeal, McPherson JA correctly said:

Whether or not [the pleading] ought to and will be struck out [as disclosing no cause of action] is ultimately a matter for the discretion of the judge who hears the application. In the end, however, it depends on the degree of assurance with which the requisite conclusion is or can be arrived at. The fact that reasonable minds may possibly differ about whether or not the material is capable of a defamatory meaning is a strong, perhaps an insuperable, reason for not exercising the discretion to strike out. But once the conclusion is firmly reached, there is no justification for delaying or avoiding that step [at] whatever stage it falls to be taken.

385 Google's application in this matter, for reasons explained at [3] to [5] above, falls to be considered in the same way.

386 The question whether the matter relied upon by the plaintiff is capable of giving rise to the defamatory imputations alleged involves considerations of capability at two points. Framed as questions, they are: (1) is the matter complained of capable of conveying the pleaded imputations;[241] and (2) are such imputations[242] capable of being defamatory?

387 As we explained at [13] above, it is necessary to consider Google's application in the context of the plaintiff's pleaded case, subject to resolving the identified ambiguity in the plaintiff's favour. In our opinion, it is quite clear, as was submitted for Google, that the plaintiff has pleaded, in the case of the images matter and the web matter, a composite publication. The reason why publication has been alleged in that form need not be speculated upon. There are cases in which an issue has arisen whether a plaintiff is entitled to rely upon composite publication. Burrows was such a case. In other instances, a defendant has attempted to force a plaintiff to 'strike in' additional material so as to create a composite publication. Cripps v Vakras[243] was such a case. Generally speaking, it has been accepted that it is for the plaintiff to choose whether he or she pleads composite or discrete publication.[244] There are limits to what may be pleaded as a composite publication, but the question whether the plaintiff's pleading in the present case travels beyond those limits was not a live matter either below or in this Court.

388 It follows from what we have said that the single meaning to be attributed to the images matter and the web matter is to be determined having regard to the entirety of the matter relied upon, not some fragment of it. That is so, although in the case of the web matter, the printouts comprising Annexure B are of three different kinds.

389 In the present case, if any of the defamatory imputations which are pleaded was arguably conveyed, it could not be doubted that it was capable of being defamatory. So the question to be determined is whether Google has established that the plaintiff has no real prospect of success in attempting to show that the matter complained of is capable of conveying any of the pleaded imputations.

390 In that inquiry, in our opinion, Google was correct to submit that the question must be determined by reference to the understanding of an ordinary reasonable user of a search engine such as the Google search engine, without which the facility to navigate the trillions of pages on the world wide web would be gravely compromised. The law constantly adapts to changing circumstances. The ordinary reasonable reader of a book or newspaper became the ordinary reasonable viewer of a television program. The internet, in its various manifestations, is a newer - though now by no means a new - vehicle for viewing printed words and images. It is quite clear that the use of search engines, and in particular the Google search engine, is ubiquitous worldwide. The capability of displayed search results to defame should be considered by reference to the ordinary reasonable user of such a site.

391 In our opinion, so approached, the plaintiff would have no prospect at all of establishing that the images matter conveyed any of the defamatory imputations relied upon. The characteristic of the 20 pages of printouts described at [25] above is very pertinent to that conclusion. So, also, in our opinion, are the considerations to which we have referred at [145] to [151] above. It might be said, if a contrary conclusion was to be reached, that the list of persons potentially defamed would be both large and diverse. We do not accept that such a conclusion would be sound.

392 We should add that the image which is at page 14 of Annexure A, an image of autocomplete predictions and other material posted on ozsoapbox.com, could not carry any of the imputations upon which the plaintiff sues. That is so for a number of reasons.

393 First, we agree with the conclusion of Blue J in Duffy that autocomplete predictions are incapable of being defamatory. As his Honour put it, an

ordinary reasonable person reading autocomplete predictions would understand that they are neither a statement by Google nor a reproduction by Google of a statement by someone else... Rather they comprise a collection of words that have been entered by previous searchers when conducting searches.[245]

His Honour might have added that the autocomplete predictions which are returned in a particular instance are strongly influenced by the particular user's previous searches.

394 Second, the image itself refers to the 'website for this image', stating that, 'in a nutshell, Michael Trkulja's beef with both Yahoo! and Google was that ...'. Also on the page are advertisements for a defamation lawyer and an online solicitor. It is, in the event, crystal clear that the image has to do with the plaintiff's earlier successful defamation proceedings.

395 Third, whilst, viewed discretely, this particular page could not be considered capable of carrying any of the pleaded imputations, that position is a fortiori when the page is considered in the entire context of the images matter comprised by pages one to 20 of Annexure A.

396 We turn to Annexure B. In our opinion, again, the plaintiff could not possibly succeed in showing that the web matter upon which he relies carried any of the pleaded defamatory imputations.

397 Focusing discretely upon pages five to seven, what we have said at [393] above is in point.

398 We turn next to page one of Annexure B. Considered discretely, it is essentially of the same character. As we earlier noted, it is a screenshot of comments made on a webpage referable to the plaintiff's earlier proceeding against Google - as can be seen from the URL at the bottom of the page. The page shows part of an autocompletion, and then mordantly refers to the plaintiff being 'Streisand'd'. The whole point of this page is that the plaintiff's successful defamation proceeding had produced the Streisand effect. Far from carrying any of the defamatory imputations pleaded by the plaintiff, the commentator was pointing out that the plaintiff's successful defamation proceedings - in which he had been awarded damages in respect of an imputation that he was somehow connected with the Melbourne underworld - had not brought matters to an end.

399 Still focusing on the pages of Annexure B discretely, we have already described the content of page two.[246] The only reference to the plaintiff is in connection with his earlier successful defamation proceeding against Google, under the heading 'Google defamation case' and 'publishing in the digital age - crikey'. How that could possibly be said to be defamatory of the plaintiff we do not understand. The fact that the reference to the plaintiff's earlier successful defamation proceeding was on a results page which adverted to the television series, 'Underbelly', which contained thumbnails of persons associated with the Melbourne underworld (none of which were the plaintiff) and which referred to a reputed criminal named Arico, could not possibly deflect attention from the import of the only reference to the plaintiff. We add only that the sample thumbnails displayed on this printout were the five images, not including that of the plaintiff, which appear in that sequence on the top line of page 17 of Annexure A. This seems unlikely to be mere coincidence, because both searches were conducted on 23 November 2012.

400 Page three is a printout of another web search conducted on 23 November 2012. We have described its content at [37] above. Having regard to the fact that the thumbnail of the plaintiff was contained in a random compilation of images which included a crime reporter, the Google logo, and text dealing with the plaintiff's earlier successful defamation proceeding, and that the only relevant snippet concerned the plaintiff's successful defamation proceeding against Yahoo!, we fail to see how anything on this page of the Annexure could possibly be defamatory of the plaintiff.

401 We turn to page four of Annexure B. We have referred to its content at [38] above. It may be regarded as the high water mark of the material relied upon by the plaintiff, because of the fact that the return of images included the plaintiff and three criminals. Pausing, and underlining the random nature of the images displayed, the four images in the particular sequence are the first four images from the left on the top line of page five of Annexure A. It is noteworthy, for reasons to which we will advert in a moment, that the date of the image search which returned page five of Annexure A is unknown, it being the only page of Annexure A which is undated; and that page four of Annexure B is also undated. It can be said, however, that the two searches were apparently conducted about 16 minutes apart, and - one might infer - very probably on the same day.

402 The heading under which the thumbnails on page four appear is 'Images for Melbourne underworld criminals - report images'. It is a similar heading to that which appears above the compilations of images on pages two and three of Annexure B. A reasonable user of the internet, aware of the unpredictable results which are generated by an image search - well exemplified by the 20 pages of Annexure A - would immediately apprehend, in our opinion, that the thumbnails on page four of Annexure B were of no different character. They could not convey the defamatory imputations pleaded by the plaintiff.

403 But there is a further matter. If the plaintiff had pursued his case as one of secondary publication, then, so far as notice may be relevant in such a context, notice was not given until 3 December 2012. It is scarcely conceivable that assumed secondary publication prior to that date would not attract a successful innocent dissemination defence. But according to the plaintiff's particularised case, it is impossible to say whether page four of Annexure B (and the same is the situation with page five of Annexure A) was published after the giving of notice. Even if the defendant carried the burden of ultimate persuasion that page four was published before notice was given, the plaintiff's pleaded case would raise no evidentiary case to the contrary.

404 Thus far, we have considered capacity to defame, in the context of the pages of Annexure B, by focusing upon each individual page discretely. But when the pages are viewed in their entirety, Google's submission that the matter complained of is not capable of making out the defamatory imputations complained of - because the ordinary reasonable user of the internet would not understand the content of the search results in such a way - is emphasised. So, for more than one reason, in our opinion, the judge erred in finding that the matter complained of was capable of being defamatory of the plaintiff; and, in particular, his Honour erred in placing store upon page four of Annexure B in the way which he did.

405 In the event, we conclude that the matter upon which the plaintiff relies is incapable of conveying any of the defamatory imputations which are pleaded.

406 The judge concluded that Google was capable of being a publisher of the material complained of by the plaintiff. We agree that, if his claim had been pursued in a way other than the way in which it was in fact pursued, this could be so.

407 His Honour further stated that if Google[247] was to have immunity from suit, it must be bestowed by the legislature. He observed that Division 2 of Part IV of the Act, which deals with a number of statutory defences, is relevantly silent. In his Honour's view, that stood against the common law now introducing search engine immunity.

408 The immunity contended for by Google is quite specific. That is, that the owner/operator of a search engine ought have immunity from liability in a defamation proceeding when liability is said to depend on publication of automatically generated materials returned by the search engine in response to an individual user's query.

409 The parties advanced submissions in writing. They did not supplement them orally.

1. The issue of whether the owner/operator of a search engine should be liable, in respect of automatically returned materials, is a novel one and difficult one. As earlier parts of these submissions demonstrate, there is no "settled rule of the common law"[248] that such a defendant is a publisher.
2. The public interest in freedom of speech (including the right to seek, receive and impart information) should be given effect in the development of the law of defamation.[249] The common law of defamation was developed by the courts; that development by the courts continues today.[250]
3. With due respect to [the judge], it cannot be correct that, in the novel circumstances presented by the Internet and search engines, the common law must nevertheless remain static, the Court must stay its hand, and members of the Australian community must wait and see whether six States and two Territories will legislatively provide. Just as the law of defamation has in the past evolved to address changes in communication media, it must now evolve to meet the challenges presented by the proliferation of Internet communications and, in particular, the role played by search engines.
4. Recognition of the immunity would protect what are, in contemporary society, essential channels of communication. Freedom of speech, including ability to research and obtain what others have written, is of fundamental importance to a democratic society. The immunity will also protect and foster growth of the digital economy, thereby increasing our society's prosperity.[251] Without protection, search engine operators will have to err on the side of caution and remove even lawful content when threatened with court action. This Court should hold that the immunity is available to search engine operators.

1. As articulated this proposition is advanced by the applicant regardless of whether or not notification has been given to the owner/operator of the search engine. Following notification at least, the proposition is self-evidently false. Again, however, the applicant is asking an appeal Court to address this most important of issues without a proper evidentiary base. Again, leave should be refused on discretionary grounds even if the applicant were to make the hurdle of "real prospect of success".
2. The reliance upon Lange v ABC in para 42 of their Written Case (A16) is misconceived. The implied constitutional freedom addresses only discussion of "governmental and political matters" - not private matters.
3. As to paras 43 and 44 (A16) such arguments do not assist the applicant here. The lack of an evidentiary base is again repeated. If the applicant's wishful thinking were to be accommodated the law of defamation in Australia would be rendered nugatory. Anyone wishing to defame a fellow citizen could use the internet in an anonymous way to do so.

412 We are in no doubt that, so far as the law of defamation is concerned, the question of publication in the varying contexts of the world wide web gives rise to problems - some of greater and some of lesser difficulty. The problem of results produced by a search engine's automated processes is an acute one. It has led to conflicting analyses in the common law world. The resolution to the issue of publication which we consider apt is not without its difficulties.

413 The great utility of search engines cannot be doubted. On the other hand, assuming that there is publication of automated search results, the prospect of a publication conveying defamatory imputations in a particular instance cannot be dismissed. A balancing of interests, in those circumstances, must be undertaken. We agree with the judge below that the balancing is a matter for Parliament, not the courts. The submission for Google that it would be too hard to get the various Australian jurisdictions to speak with one voice cannot deflect attention from what is the correct, rather than the expedient, course.

414 In the United States, with the emphasis in that country on free speech, legislation has, to an extent, addressed the problem of publication on the internet.[252] The legislation has produced, on occasion, an unhappy result: Carafano v Metrosplash.com Inc.[253] The American experience suggests that the content of any Australian legislation would require much thought. But one thing, in our opinion, is clear. If there is to be any immunity in favour of a search engine from liability for defamation, it must be conferred by legislation.

415 Google should have leave to appeal against the orders made below on 17 November 2015, and a consequential costs order made on 30 November 2015. The appeal should be allowed. In lieu, it should be ordered that the plaintiff's amended writ and amended statement of claim dated 13 June 2014 and their purported service on Google Inc be set aside.

[2] [2000] HCA 41; (2000) 201 CLR 552 ('Agar').

[5] See Civil Procedure Act 2010 r 64.19.

[6] See Supreme Court Act 1986 s 14C and Kennedy v The Shire of Campaspe [2015] VSCA 47, [13] (Whelan and Ferguson JJA).

[7] It does not deny the problems of self-representation, but it is the fact that the plaintiff has been plaintiff in a good deal of other defamation litigation, and, in the course of doing so, has at times prepared his own pleadings and at times appeared unrepresented: Trkulja v Yahoo! Inc [2010] VSC 215; the related matters of Trkulja v State of Victoria No 4187 of 2009 and Trkulja v Google Inc [2010] VSC 226; see also Trkulja v Google Inc (No 5) [2012] VSC 533; Trkulja v Markovic [2013] VCC 1095 and, on appeal, Trkulja v Markovic [2015] VSCA 298; Trkulja v Dobrijevic [2016] VSC 421 and [2016] VSC 596. He has engaged in other litigation also, at times self-represented. A sample was referred to by Beach J in Trkulja v Google Inc (No 4) [2011] VSC 560, [18] and n 5.

[8] A judge had given the plaintiff leave to file an ASoC only.

[9] [2012] VSC 533 ('Trkulja (No 5)'). The proceeding to which Beach J's ruling relates will be referred to as 'the earlier Google proceeding').

[10] The 'First' defendant is a reference to Google. The 'Second' defendant refers to Google Australia, which is no longer a party to the proceeding.

[11] Or by some person having a community of interest with the plaintiff. For convenience, we will not keep repeating this alternative. We are, of course, conscious that publication of defamatory material to a plaintiff only is no publication, and we do not mean to imply that others did not access the search results which are complained of.

[12] A copy of Annexure A is Annexure 1 to these reasons. The pages are sequenced in the same order in these reasons and in Annexure 1. A reference in these reasons to a page number in Annexure A is to be understood as a reference to that page number of Annexure 1.

[14] Of the 19 instances where the search date is detectable, five (pages six, 14, 15, 17 and 18) predate 1 December 2012, and thus predate the concerns notice dated 3 December 2012.

[15] Paragraph five of the original statement of claim.

[16] Or 'convicted "Melbourne underworld criminal''', or the like.

[17] Fixed by the time when the website displaying those images had last been 'crawled' and 'indexed' by the search engine. See the explanation of these terms later in these reasons.

[18] The 20th page (which is page 14 of Annexure A), is an image of an autocomplete prediction, posted on ozsoapbox.com/wp.

[23] Including Ms Jackie Weaver and Messrs Joel Edgerton and Ben Mendelsohn.

[24] Trkulja v Google Inc [2015] VSC 635, [15] ('Reasons').

[25] That is, for leave to appeal; and, if leave be granted, the substantive appeal.

[26] The plaintiff pleads both false and true innuendos. The false innuendo plea would seem destined to fail if the search term was not part of the allegedly defamatory matter.

[27] A copy of Annexure B is Annexure 2 to these reasons.

[29] According to Wikipedia, 'the phenomenon whereby an attempt to hide, remove or censor a piece of information has the unintended consequence of publicizing the information more widely, usually facilitated by the internet': Wikipedia, 'Streisand Effect' (13 December 2016) . The term was coined in 2005 'in relation to a holiday resort issuing a takedown notice to urinal.net (a site dedicated to photographs of urinals) over use of the resort's name.' All that the takedown notice did (as with Barbara Streisand's earlier attempt to suppress photographs of her residence in Malibu) was that more people than might have been expected viewed the resort's urinal depicted on the website.

[32] The date of the search is not revealed.

[33] They are respectively dated 15 June 2013, 24 January 2013 and 24 January 2013.

[34] Annexed to the notice were printouts of a number of the pages which appear in Annexures A and B to the ASoC, and also printouts of pages from, inter alia, 'YouTube PinkP4nter's Channel';'knightsarmy86976.yuku.com/topic...';'sabotagetimes.com/life/the...';'ozsoapbox.com/Melbourne/...'; 'ozsoapbox.com/rest-of-australia/...'; and 'ozsoapbox.com/melbourne/ a-curious...'.

[35] Although in 2007 the plaintiff apparently claimed to know the person's identity.

[37] Reasons [9], referring to Trkulja v Google (No 5) [2012] VSC 533, [27].

[38] [2014] HKLRD 493, [105] ('Yeung').

[43] [2006] EWHC (QB) 407; [2006] 3 All ER 336 ('Bunt').

[44] [2011] 1 WLR 1743 ('Metropolitan Schools').

[45] [2012] EWHC 449 and, on appeal[2013] EWCA Civ 68; , [2013] 1 WLR 2151 ('Tamiz').

[46] [2014] NSWSC 897; (2014) 88 NSWLR 670 ('Bleyer').

[50] [1928] HCA 50; (1928) 41 CLR 331 ('Webb').

[51] [1996] HCA 38; (1996) 186 CLR 574 ('Thompson').

[52] [2013] HKCFA 47; [2013] 5 HKC 253 ('Fevaworks').

[53] [2015] SASC 170; (2015) 125 SASR 437 ('Duffy').

[54] Reasons [54]-[56] (citations omitted).

[62] [1928] HCA 50; (1928) 41 CLR 331, 364 (citations omitted).

[63] But see now the possible operation of s 32 of the Defamation Act 2005.

[64] Paraphrasing Michael Gillooly, The Law of Defamation in Australia and New Zealand (Federation Press, 1998).

[65] [1937] 1 KB 818 ('Byrne').

[66] (1991) Aust Torts Reports 81-127 ('Urbanchich').

[67] Byrne [1937] 1 KB 818, 830.

[70] (1991) Aust Torts Reports 81-127, 69,193.

[72] (1885) 16 QBD 354 ('Emmens').

[74] [1900] 2 QB 170 ('Vizetelly').

[77] [1996] HCA 38; (1996) 186 CLR 574.

[86] See Matthew Collins, Collins on Defamation, (Oxford University Press, 1st ed, 2014), variously at [16.30]-[16.59].

[87] The analysis of material in the two annexures to the ASoC by Vaughn Beckett Madden-Woods in his affidavit affirmed 19 December 2004 and filed on behalf of Google, although the material served on Google did not wholly correspond with the annexures to the ASoC as filed, shows that most of the searches were made on the Google.com.au website, a few on the Google.com website, and one on the Google.com.de website. The last-mentioned is the country-specific Google website for Germany. This correlates with paragraph 13 of the ASoC, which identifies four of the persons to whom the material was allegedly published as being Australian residents, one as being resident in Germany, and one as being resident in Banja Luka, Bosnia.

[91] These examples do not deal with the situation of search words separated by hyphens.

[92] For instance, Matthew Collins, Collins on Defamation, (Oxford University Press, 1st ed, 2014); Alistair Mullis and Richard Parkes et al (eds), Gatley on Libel and Slander (Sweet & Maxwell, 12th ed, 2013); and, very recently, Jaani Riordan, The Liability of Internet Intermediaries (Oxford University Press, 2016).

[93] [1997] USSC 73; 521 US 844.

[94] International Telecommunication Union, ICT Facts and Figures 2016 (June 2016); see also Riordan, above n 92, [1.13].

[95] Australian Bureau of Statistics, Household Use of Information Technology, Australia, 2014-15 (February 2016).

[96] See also, Herscovici affidavit [13].

[97] Riordan, above n 92, [2.24] (citing Barry Leiner et al 'Brief History of the Internet' (1995) ).

[98] As at 2014, in the United Kingdom, it was reported that 66 per cent of internet users had at least one social networking account: Riordan, above n 92, [2.61] n 82.

[99] Riordan, above n 92, [2.47] (citing Barry Greene and Philip Smith, Cisco ISP Essentials (Cisco Press, 2002), 229-234).

[100] A standardised publication format which allows end users to access syndicated feeds of media or data, commonly used to subscribe to blogs.

[102] (2013) 249 CLR 435 ('ACCC').

[107] Madden-Woods affidavit [63]-[64].

[108] Madden-Woods affidavit [65]-[66].

[111] Madden-Woods affidavit [75]-[76].

[112] Madden-Woods affidavit [77], [80].

[113] Madden-Woods affidavit [81]-[82].

[116] Madden-Woods affidavit [96]-[98].

[118] Madden-Woods affidavit [102]-[105].

[119] Madden-Woods affidavit [106]-[108].

[120] Madden-Woods affidavit [111]-[114], [117]-[120].

[122] Madden-Woods affidavit [136], [138]-[140].

[124] Madden-Woods affidavit [131]-[132], [155], [157]-[164].

[126] [2012] VSC 533, [18].

[128] [1928] HCA 50; (1928) 41 CLR 331, 363-364.

[130] [2011] 3 SCR 269 (Supreme Court of Canada) ('Crookes').

[131] [2015] SASC 170; (2015) 125 SASR 437.

[132] [2013] HKCFA 47; [2013] 5 HKC 253.

[133] [2014] NZCA 461; [2014] 3 NZLR 722 ('Murray').

[134] See, variously, transcript 46-59 of the hearing in this Court. At one point, plaintiff's counsel foreshadowed an application to amend the ASoC to insert a new paragraph 19 as follows: 'Notwithstanding the matters pleaded in paragraphs 16 to 18 hereof, the defendant continued to publish the Google images matter and the Google web matter from on or about 20 December 2012 to the date hereof'. The application was not pursued. Perhaps, in conjunction with the existing paragraph 26, the foreshadowed amendment would have sufficiently raised a post-notification case.

[135] Transcript 67, lines 25-31 to transcript 68, lines 1-9.

[136] Either the parent company or the Google entity in a particular country.

[137] See also Anne S Y Cheung, 'Defaming by Suggestion: Searching for Search Engine Liability in the Autocomplete Era' (Research Paper No 2015/018, University of Hong Kong Faculty of Law, 2015).

[138] [2006] EWHC (QB) 407; [2006] 3 All ER 336.

[140] His Lordship had already held that there was no sustainable case to support the proposition that one of the ISPs had knowingly authorised, sanctioned or participated in any of the relevant publications: ibid 345 [35]. To that extent, the broader proposition was probably unnecessary.

[141] Metropolitan Schools [2011] 1 WLR 1743; Tamiz [2013] EWCA Civ 68; [2013] 1 WLR 2151.

[142] [2012] VSC 533, [29]. His Honour cited Urbanchich (1991) Aust Torts Reports 81-127 in that connection.

[143] Matthew Collins, The Law of Defamation and the Internet (Oxford University Press, 3rd ed, 2010), [6.45]-[6.50] .But compare the author's more circumspect treatment in Collins, Collins on Defamation, above n 92, [4.62]-[4.65].

[144] Mullis and Parkes et al, above n 92, [6.27]; Riordan, above n 92, [8.90]-[8.95].

[145] [2013] EWCA Civ 68; [2013] 1 WLR 2151, 2161 [23] (Richards LJ).

[146] [2014] NSWSC 897; (2014) 88 NSWLR 670, 684 [77].

[148] This analysis is preferred by Mullis and Parkes et al, above n 92, [6.47].

[149] [2013] EWCA Civ 68; [2013] 1 WLR 2151, 2162 [25].

[152] Murray v Wishart [2014] NZCA 461; [2014] 3 NZLR 722, [170].

[158] Ibid.

[160] [2015] SASC 170; (2015) 125 SASR 437, [180].

[161] [2013] HKCFA 47; [2013] 5 HKC 253, [33].

[162] Ibid [46]-[49] (citations omitted).

[163] Ibid [50]-[54] (citations omitted).

[164] And with single judge decisions cited in Tamiz [2013] EWCA Civ 68; [2013] 1 WLR 2151.

[165] Fevaworks [2013] HKCFA 47; [2013] 5 HKC 253, [76].

[166] Ibid [84] (citations omitted).

[169] [2001] QB 201 ('Godfrey').

[170] [2012] 3 CMLR 104 ('Davison'). See Tamiz [2013] EWCA Civ 68; [2013] 1 WLR 2151, 2163-2164 [30]-[31].

[171] [2014] NZCA 461; [2014] 3 NZLR 722, [118].

[172] (2016) NIQB 7, [67].

[174] The Fevaworks analysis is strongly endorsed by Mullis and Parkes et al, above n 92, [6.29]. But Riordan gives the Byrne analysis (highly qualified) support: Riordan, above n 92, [8.73]-[8.83].

[175] [2012] NZHC 2352 ('A v Google NZ').

[176] [2015] BCSC 1024 ('Niemela').

[177] Metropolitan Schools [2011] 1 WLR 1743, [42].

[181] The hyperlink was returned on both the images and webpage results.

[182] Including the underlying article.

[183] Again, including the underlying article.

[184] "In Cassell & Co Ltd v Broome [1972] UKHL 3; [1972] AC 1027 at 1085, Lord Reid warned of the danger of placing reliance on the literal words of particular judgments instead of searching for the applicable principle. His Lordship said:

'... experience has shown that those who have to apply the decision to other cases and still more those who wish to criticise it seem to find it difficult to avoid treating sentences and phrases in a single speech as if they were provisions in an act of Parliament. They do not seem to realise that it is not the function of noble and learned lords or indeed any judges to frame definitions or to lay down hard and fast rules. It is their function to enunciate principles and much that they say is intended to be illustrative or explanatory and not to be definitive'."

Cited in Cain v Glass (No 2) (1985) 3 NSWLR 230, 248 (McHugh JA).

[185] [2006] EWHC (QB) 407; [2006] 3 All ER 336.

[186] [2011] 1 WLR 1743.

[187] [2012] EWHC 449.

[188] See generally, and further, the discussion about intent by Hunt J in Urbanchich (1991) Aust Torts Reports 81-127, wherein his Honour said:

The law of defamation ... has never required 'a conscious intent to induce the public or any individual to read the alleged libels' ... .

The proposition that conduct of a passive nature cannot amount to publication ... was decisively rejected in Byrne v Deane ... . Greene LJ said he was quite unable to accept any such proposition. Nor am I. ...

There will, of course, always be issues (which will usually be for the jury to determine) as to whether the defendant has the ability to remove the defamatory statement and whether the time given before the commencement of the action was a reasonable one in which to do so. ...

It is clear from all of those authorities that the facts upon which the plaintiff relies here - notice of the existence of the defamatory statement, an ability to remove it [or in the present case block it] and the failure to comply within a reasonable period where the request to do so - may, if accepted by the jury (in particular, whether the period given was reasonable in the circumstances), give rise to the required inference that the defendant had in fact accepted a responsibility [which can be accepted by showing the defendant consented] for the continued publication ...

[189] Webb [1928] HCA 50; (1928) 41 CLR 331; Urbanchich (1991) Aust Torts Reports 81-127.

[190] Trkulja (No 5) [18]-[19], [28]-[31] (citations omitted).

[191] Briefly, at [2013] EWCA Civ 68; [2013] 1 WLR 2151, 2161 [21], as part of an attack on the correctness of Bunt.

[192] [2012] NZHC 2352, [73].

[193] [2014] NSWSC 897; (2014) 88 NSWLR 670, 682-683 [66]-[72]; 684-685 [78]-[79].

[198] See [2014] HKLRD 493, [103], [106].

[199] See [2015] SASC 170; (2015) 125 SASR 437, 475-479 [170]-[180].

[201] Ibid 495-496 [203]-[209] (citations omitted).

[202] Ibid 496 [211]-[212] (citations omitted). We have left open the question whether the Byrne or Emmens line of authority is to be preferred in the case of a website forum provider. Nothing turns on it for present purposes.

[204] Ibid 500-501 [232]-[234] (citations omitted).

[205] [2015] SASC 170; (2015) 125 SASR 437, [373]-[376].

[206] Google had blocked the URLs from appearing on returns from searches on its Google.ca site. But the plaintiff sought to compel blocking on the Google.com site.

[207] [2015] BCSC 1024, [22].

[209] Ibid [86]-[108] (citations omitted).

[210] Though not always in the case of the content of hyperlinked webpages.

[211] That line of authority, however, is, in our view, relevant in the internet forum host cases.

[212] Y_eung_ [2014] HKLRD 493 and Duffy [2015] SASC 170; (2015) 125 SASR 437.

[213] Or is reckless or careless as to such communication occurring in consequence of his or her conduct.

[215] Mullis and Parkes et al, above n 92, [6.27].

[216] [2012] EWHC 449, [39].

[217] [2013] EWCA Civ 68; [2013] 1 WLR 2151, [23].

[218] [2010] 3 All ER 560, [23].

[220] [2015] BCSC 1024, [94].

[221] Canada Act 1982 (UK) c 11, sch B pt I.

[224] As, it may be, an algorithm may be designed.

[225] Whether at common law, or under s 32 of the Act. The common law and statutory defences run side by side: see s 24(1) of the Act.

[226] Reasons [17]. But this may have been directed to the plaintiff not pleading republication of any hyperlinked material.

[229] Even so, that would not be a basis for granting leave to appeal and allowing the appeal if the correctness of the order was not successfully impugned. See [97] above.

[230] We have not ignored the fact that, at one point, plaintiff's counsel submitted that Thompson was 'really of no assistance at all'. But that was directed to the reliance by Google's counsel on so much of Thompson as held that the secondary publisher/innocent dissemination principles may be called in aid by an electronic publisher - always depending on the circumstances.

[231] The position could be no different if, as plaintiff's counsel insisted was necessary, there was cross-examination at trial.

[232] So far as it relates to publication of the 'images matter', see [17]-[21] and [24]-[26] above. So far as it relates to the 'web matter', see [31]-[42] above.

[238] Favell v Queensland Newspapers Limited [2005] HCA 52; (2005) 221 ALR 186 ('Favell').

[239] (1987) 10 NSWLR 651 ('Burrows').

[240] [2005] HCA 52; (2005) 221 ALR 186, 189 [6].

[243] [2014] VSC 110.

[244] See Phelps v Nationwide News [2001] NSWSC 130, [22].

[245] Duffy [2015] SASC 170; (2015) 125 SASR 437, [375].

[247] Or it could be any search engine proprietor.

[248] State Government Insurance Commission v Trigwell [1979] HCA 40; (1979) 142 CLR 617 at 633 (Mason J).

[249] Lange v Australian Broadcasting Corporation [1997] HCA 25; (1997) 189 CLR 520; Reynolds v Times Newspapers Ltd [2001] 2 AC 127; Jameel v Wall Street Journal Europe Sprl [2006] UKHL 44; Grant v Torstar Corp [2009] 3 SCR 640.

[250] The Uniform Defamation Acts do not constitute a codification of the law of defamation. See generally: section 6 of the Defamation Act 2005 (Vic); Setka v Abbott [2014] VSCA 287.

[251] See expert report of Dr Richard Mark Simes.

[252] 47 USC 230.

[253] [2003] USCA9 524; 339 F 3d 1119 (9th Cir, 2003).