When AI says something false about a business: what the cases show so far

When an AI answer says something false about a business, an obvious question is whether anyone can be made to answer for it. A handful of public cases now give a partial answer. It is mixed, slow and still moving.

This is a summary of public decisions and filings as of 27 September 2026, written for people who run companies. It is not legal advice. Air Canada is not a Miscited customer, and neither is any other company named here.

The cases at a glance

A company answers for its own chatbot

Moffatt v. Air Canada, 2024 BCCRT 149, is the only case here about a company's own bot. A customer asked Air Canada's chatbot about bereavement fares and got an answer that the policy page it linked to contradicted. Air Canada argued that it could not be held liable for information provided by its agents, including a chatbot. The tribunal called that "a remarkable submission" (para 27) and ordered the airline to pay C$812.02 in total (para 44). We went through the decision in what the Air Canada chatbot ruling means for what AI says about your company.

The practical reading is the plain one: if the bot is yours, treat its answers as yours. The rest of this page is about assistants a business does not run, where the position is much less settled.

Walters v. OpenAI: a false answer nobody believed

Mark Walters, a radio host, sued OpenAI after ChatGPT gave a journalist a false summary of a lawsuit brought by the Second Amendment Foundation, saying Walters had been accused of embezzling from it. On 19 May 2025 the Superior Court of Gwinnett County in Georgia granted summary judgment to OpenAI. As summarised by Loeb & Loeb and by Cleary Gottlieb, it gave three separate grounds.

1. No defamatory meaning. A reasonable reader in the journalist's position could not have concluded that the output stated actual facts. ChatGPT had said it could not access the internet and had a knowledge cutoff, OpenAI's terms warned that its output can be inaccurate, and the journalist established within about an hour and a half that the summary was untrue.
2. No fault. There was no evidence of ordinary negligence, and the court held that Walters was at least a limited-purpose public figure who had to show actual malice, which the evidence did not support.
3. No damages. Walters conceded that there had been no harm. Punitive damages were not available because nobody had asked OpenAI for a correction or retraction, as Georgia law requires.

For a business, the reasons matter more than the result. The result rested on unusual facts: one reader, who did not believe the output and did not pass it on. The more useful detail is the last one. Punitive damages failed because nobody had asked OpenAI to correct the output. Asking for a correction, and keeping a record that you asked, comes up again in the cases below.

AI Overviews: the claims still open

Two suits in the United States over Google's AI Overviews are still open.

Wolf River Electric

LTL LED, LLC, a Minnesota solar installer trading as Wolf River Electric, sued Google in Minnesota state court in March 2025, together with four of its executives. The complaint alleges that AI Overviews told searchers the Minnesota Attorney General was suing the company over deceptive sales practices, that no such lawsuit existed, and that the pages Google cited did not support the statement:

"Google cited numerous sources in support of its false assertions; however, none of the referenced materials in fact contained the information Google claimed they did."

The complaint describes a customer ending a contract with a total price of $150,000 because of what Google published (paras 175 to 176), and the plaintiffs' initial disclosures estimated damages at about $110 million to $210 million. These are allegations and estimates, not findings.

So far the case has turned on procedure. Google moved it to federal court on 9 June 2025. On 9 January 2026 the federal court held that the move came too late, because a letter served with the complaint in March 2025 had already claimed $24.7 million in damages for 2024, and sent the case back to Ramsey County District Court. We could not find a ruling on the merits in the public record as of 27 September 2026.

Keene v. Google

Author Jimmy Keene alleges that at least four AI Overviews, between about May and June 2025, made false statements about Keene, including one that described Keene as serving a life sentence without parole. Each cited a Wikipedia page that, according to the complaint, did not contain the statements.

On 14 September 2026, Judge Thomas M. Durkin of the Northern District of Illinois denied Google's motion to dismiss the claims over two of those AI Overviews. The court dismissed a third as substantially true and a fourth because publication to anyone else was not alleged, both without prejudice. Two points in the opinion matter to any business.

• A citation does not turn a statement into an opinion. In the court's words: "The mere act of providing a citation to an assertion does not convert that assertion from fact to opinion." (Memorandum Opinion and Order, p. 5.)
• Notice matters. Keene alleges three complaints to Google, and Google's own exhibit showed a legal removal request submitted on 25 May 2025. Either someone at Google read it, the court reasoned, which could support knowledge of falsity, or nobody did, which could support reckless disregard. Actual malice was plausibly alleged.

The court also found Google's reliance on Walters unpersuasive. Asking an AI tool to produce work and getting several signs that the output may be false, it reasoned, is different from a basic Google search, where an ordinary user is looking for factual information rather than AI work product.

Munich

Outside the United States, the Munich Regional Court preliminarily ruled that Google is liable for false statements in AI Overviews that linked two publishers to scams and dubious business practices, with an order requiring a temporary injunction, according to reports by Ars Technica and WIRED in June 2026. Google had not corrected the output even after a cease-and-desist letter, and the court reasoned that only Google could correct it. A Google spokesperson suggested the decision could be appealed.

What the cases show so far

• Your own bot speaks for you. Moffatt is the clearest signal, and it points one way.
• For assistants you do not run, the law is unsettled. Walters lost on facts specific to one reader. Keene and the Munich publishers have got further, at early stages. In the United States cases here, no court has yet found an AI company liable.
• Notice and correction requests keep coming up. Walters could not seek punitive damages without having asked for a correction. Keene's removal request supported the claim of actual malice. In Munich, the publishers had sent a cease-and-desist letter, and the court stressed that only Google could correct the output.
• It is slow. Wolf River was served in March 2025 and, more than 18 months later, has no ruling on the merits that we could find. Keene took more than a year from filing to get past a motion to dismiss.

The practical path

None of this makes a lawsuit a plan. The claims that have survived took more than a year to get past the first round, and the buyers who read the false answer made their decisions long before that.

The path you control is shorter: find the false claims, trace each one to the page it came from, correct that page or ask its publisher to, and test whether the answer changed. Keep dated records of what the answers said and of every correction you asked for. The cases above suggest those records matter if a dispute ever does reach a court.

We set out the checking method in how to check what ChatGPT, Gemini and Perplexity say about your company and the fixes in how to fix a wrong AI answer about your company. Before acting on a single transcript, read how many prompts an AI visibility number actually needs. To see whether there is anything to fix, the free answer audit is a quick way to start.

Sources

1. Moffatt v. Air Canada, 2024 BCCRT 149, 14 February 2024, paras 27 and 44.
2. Walters v. OpenAI, L.L.C., Loeb & Loeb, May 2025.
3. Georgia Court Dismisses Defamation Lawsuit Against OpenAI Over ChatGPT Output, Cleary Gottlieb, 27 May 2025.
4. Plaintiffs' Joint Amended Complaint, LTL LED, LLC v. Google LLC, filed in the federal case as Doc. 11-13, No. 25-cv-2394 (D. Minn.).
5. Order granting the motion to remand, LTL LED, LLC v. Google LLC, No. 25-cv-2394 (D. Minn.), 9 January 2026. Federal docket.
6. Large Libel Models: Small Business Sues Google, Claiming AI Overview in Searches Hallucinated Attorney General Lawsuit, Eugene Volokh, The Volokh Conspiracy, 11 June 2025.
7. Google Missed Key Deadline in Suit Alleging Google's AI Libeled Business, Court Holds, Eugene Volokh, The Volokh Conspiracy, 12 January 2026.
8. Memorandum Opinion and Order, Keene v. Google LLC, No. 1:25-cv-11431 (N.D. Ill.), 14 September 2026. Federal docket.
9. Defamation Claims Cleared to Proceed Against Google's AI Overviews, Federal Judge Rules, Law.com, 14 September 2026.
10. Nobody needs AI to search the Internet, court says in ruling against Google, Ashley Belanger, Ars Technica, 10 June 2026.
11. A Court Has Ruled That Google Is Liable for False Statements Generated by AI Overviews, WIRED, 13 June 2026.