Defamation on social media: posts, comments, shares, group chats and reviews

Most defamation disputes now begin online. A Facebook post about a tradesperson, a comment in a community group, an Instagram story, a thread on X, a TikTok video or a message to a WhatsApp group can all be defamatory, and the law treats them in the same way as a newspaper or a broadcast. What changes online is the scale, the speed, the permanence and the difficulty of knowing who read what.

This article covers who is responsible for social media content, how courts decide what a post means, how the serious harm element applies to small audiences, what evidence matters, what to do about deletion, and what courts have awarded. It is written for both the person written about and the person who wrote the post.

Posts, comments, shares and likes: what counts as publication

Defamation requires publication, which simply means communicating the matter to at least one person other than the plaintiff. The High Court in Fairfax Media Publications Pty Ltd v Voller confirmed that any person who intentionally participates in the process by which matter is made available to a third party is a publisher, regardless of whether they knew the content was defamatory.¹ That test captures most social media activity.

An original post, a comment, a reply, a story, a video and a caption are all publications by the person who wrote or uploaded them. Sharing or reposting another person’s content is, on the Voller test, an intentional and voluntary act of participation that makes the content available to a new audience, so the person who shares should expect to be treated as a publisher whether or not they add words of their own. No Australian appellate court has yet ruled on a bare share or repost, so the point has not been finally settled.¹ Quoting a post in order to criticise it still republishes the quoted words, and the person quoting may be liable for them.

Hyperlinks are different. In Google LLC v Defteros the High Court held that providing a hyperlink in a search result, without more, is not participation in the publication of the material at the other end of the link; a hyperlink is a reference to something somewhere else.² Gageler J added that a link can become publication when combined with other factors, such as words that entice or encourage the reader to follow it, which is a question of fact and degree.² A post that links to an article and summarises its sting in defamatory terms will therefore be a publication of that sting even if the link alone would not be.

Likes and reactions remain uncertain. In Bolton v Stoltenberg, Payne J declined to treat a ratepayers’ association chair’s “liking” of 64 Facebook posts as making her a publisher of them on the limited evidence before the court, but left open the possibility that a like which pushed the post into other users’ feeds could amount to participation.³ The same defendant was held liable for a comment urging others to like a defamatory post, because the comment specifically endorsed it and drew attention to it.³ Tagging a person or sharing a post into a group is more clearly participation. Anyone relying on the “like” distinction should assume it may not survive better evidence about how a platform’s algorithm works.

Page administrators and group admins

Voller concerned media companies that operated public Facebook pages on which readers commented. The High Court held, by majority, that by creating the pages and posting content that invited comments, the companies facilitated and encouraged the comments and were publishers of them, even before they knew the comments existed.¹ The principle is not confined to media companies. A person who runs a community page, a buy swap sell group or a neighbourhood group is in the same position in relation to comments by members.

The 2024 to 2025 reforms respond to that exposure. In every jurisdiction except Western Australia and South Australia, s 31A of the Defamation Act 2005 provides a defence for a “digital intermediary”, defined as a person other than the author, originator or poster who provides or administers the online service through which the matter is published.⁴ The defendant must prove that it had an accessible complaints mechanism when the matter was published, such as an email address or a messaging address, and that if the plaintiff gave a written complaint identifying the plaintiff, the matter and its location, and stating that the plaintiff considered it defamatory, reasonable steps to remove or block access were taken within seven days.⁴ The defence is defeated only if the plaintiff proves that the defendant was actuated by malice in establishing or providing the online service through which the matter was published.⁴ Queensland’s provisions, which commenced on 26 December 2025, use the same section numbers as New South Wales.⁵

For an administrator that means having a visible way for people to complain and acting on complaints within the seven day period. A s 31A complaint is separate from the concerns notice required before proceedings can be started.

How courts read social media

The test for meaning is what ordinary reasonable people would understand by the matter, considering the publication as a whole. The High Court in Trkulja v Google LLC described the ordinary reasonable person as someone who views the publication casually, is prone to a degree of loose thinking, reads between the lines in the light of general knowledge, draws derogatory implications more freely than a lawyer, and takes into account the emphasis given by conspicuous headlines or captions.⁶ Applied to social media, the question is how a casual user scrolling a feed would take a post, including its images, emojis, hashtags and the surrounding thread.

Emojis can carry meaning. In Burrows v Houda, Gibson DCJ held that a tweet reply consisting of a “zipper mouth face” emoji, in response to a question about what had happened to a solicitor referred for misconduct, was reasonably capable of conveying that the solicitor had suffered an adverse outcome, and that the words “tick tock” with a clock emoji were capable of suggesting her time was up.⁷ Her Honour observed that emojis are capable of conveying standardised meanings and have their own dictionary, while stressing that meaning remains a matter of broad impression.⁷

Context matters, and a bare tweet or headline can be defamatory even where a linked article is not. In Hockey v Fairfax Media Publications Pty Ltd, White J held that the tweet “Treasurer Hockey for sale” conveyed an imputation of corruption which the full article, read as a whole, dispelled; the tweets attracted $80,000 in damages even though the articles were not defamatory.⁸

Small groups and the serious harm element

Outside Western Australia a plaintiff must prove that the publication has caused, or is likely to cause, serious harm to their reputation.⁹ The judge decides that question, often before trial.⁹ The New South Wales Court of Appeal in Rader v Haines explained that serious harm is more than merely substantial though it need not be grave; it is not satisfied by the gravity of the imputation alone or by hurt feelings, and a grave allegation may not cause serious harm where it is published to a small number of people well acquainted with the plaintiff who do not believe it.¹⁰

That is the central problem in private group and direct message cases. In Zimmerman v Perkiss, a Facebook Messenger exchange accusing a former employee of theft was sent to one person, her new employer; the employer did not think less of her, there was no grapevine effect, and the claim was dismissed after a separate serious harm hearing.¹¹ In Rader v Haines itself, an email from parents in law to the plaintiff’s parents in England making allegations of violence was held not to have caused serious harm, because the recipients did not believe it and any effect dissipated within weeks. The email was in any event protected by qualified privilege as a communication within a family.¹⁰

At the other end of the scale, public posts by people with large followings readily satisfy the element. In Martin v Najem, Instagram videos by a food blogger calling a rival a paedophile and a racist, published to accounts with thousands of followers and accompanied by a call to followers to act, were held to have caused serious harm to each publication, with evidence of a grapevine effect, damage to the plaintiff’s health and refusals of work.¹² Gibson DCJ said that serious harm requires “fact-rich proof of harm which is actually or likely to be serious”, rather than inferences drawn from the size of the audience alone.¹²

The grapevine effect, the tendency of defamatory allegations to spread beyond the original audience, is relevant to both serious harm and damages, but it must be shown by evidence rather than assumed.¹⁰ Within closed groups, a defendant may also be able to rely on qualified privilege where the members share a legitimate common interest in the subject, although that defence is lost if the publication goes wider than the interest justifies or is actuated by malice.

Evidence: capturing and proving the post

Social media evidence disappears. Whoever is affected by a post should capture it immediately: full screenshots showing the account name, date and time, text, images and URL; the visible reactions, comments and shares; and the privacy setting or group membership that shows who could see it. Platform analytics (reach, views, impressions) can later be obtained from the account holder or by subpoena, and in Bolton v Stoltenberg own records of reach and engagement were treated as cogent evidence from which readership could be inferred.³ Publication to at least one person who actually read the material must be proved, so evidence of real readers, including people who contacted the plaintiff about the post, should be kept.

A concerns notice must specify where the matter can be accessed and, if practicable, attach a copy, so screenshots and links are needed before any letter is sent.¹³ Where the poster is anonymous, the court’s preliminary discovery rules can be used to obtain identifying information from a platform or other intermediary, and in jurisdictions that have enacted s 23A the court must weigh privacy, safety and other public interest considerations before making such an order.¹⁴

Deleting a post after a complaint

A person who has posted something they regret is often advised to take it down, and that is usually right. Removing the post limits ongoing publication, reduces the harm and is relevant to mitigation of damages. Under s 31A, taking reasonable access prevention steps within seven days of a complaint is the core of the defence for administrators.⁴

Deletion is not the same as destruction of evidence. Once a dispute is on foot, both sides are expected to preserve relevant material. A party who deletes posts, messages or analytics so that they cannot be produced in court risks adverse inferences and, in serious cases, costs consequences or contempt. The sensible course is to screenshot and export the content before removing it from public view, and to keep the export. Nobody should edit a screenshot, backdate a message or create a document after the event.

What courts have awarded

Damages in social media cases range widely. At the modest end, the commenter in Bolton v Stoltenberg who endorsed a defamatory post was ordered to pay $10,000. The author and page operator was ordered to pay $100,000, including $20,000 in aggravated damages, for a campaign of posts about a shire mayor. Both awards stood on appeal.³¹⁵ The two tweets in Hockey attracted $80,000.⁸ In Martin v Najem the court awarded $300,000 including aggravated damages for two Instagram videos accusing the plaintiff of paedophilia and racism, together with a permanent injunction.¹² Awards rise with the gravity of the allegation, the size of the audience, repetition and refusal to apologise, and fall with a small audience, prompt removal and apology.

Practical steps for each side

For a person who believes they have been defamed: capture the evidence as described above; identify everyone who published (author, sharers, page administrators); consider whether the audience and consequences are likely to satisfy serious harm; send a s 31A complaint to any administrator or platform; and, if proceedings are contemplated, send a compliant concerns notice within the one year limitation period. A measured private approach to the author often achieves removal and an apology faster than litigation.

For a person who has posted: do not respond with further posts; preserve the content and then remove it or restrict its audience; consider an apology, which is not an admission of liability under the Act; and treat any concerns notice seriously, because a reasonable offer to make amends within 28 days can be a complete defence. Where the post expressed a genuine opinion on a matter of public interest, or was true, or was made to a limited group with a legitimate interest in the subject, a defence may be available, subject to conditions that need to be checked against the facts.

Differences between jurisdictions

In Western Australia there is no serious harm element, so the small audience cases described above would be approached differently, and there is no s 31A defence; administrators must rely on innocent dissemination and the common law. South Australia enacted only part of the digital intermediary package on 15 December 2025 (a preliminary discovery provision and a power to make orders against non party intermediaries) and has no equivalent of the s 31A defence or the exemptions.¹⁶ The ACT and Northern Territory have equivalent provisions under different section numbers in the Civil Law (Wrongs) Act 2002 (ACT) and the Defamation Act 2006 (NT). Everywhere else, including Queensland since 26 December 2025, the New South Wales numbering used in this article applies.⁵

Getting advice

Whether a social media dispute is worth pursuing or defending turns on the platform, the audience, the words, the evidence of what happened afterwards and the time since publication. Advice taken early, while the evidence is still available and the limitation period is comfortably open, is usually more useful than advice taken after the thread has been deleted.