Google’s AI Overviews Defamation Ruling Is a Documentation Case

A federal judge let a libel-by-AI claim against Google move forward. The reason matters: Google was told, repeatedly, that the information was wrong.

The short version

• Google was told the information was wrong more than once. Yet similar false information continued to appear.
• Google has not been found liable. This is an early-stage ruling on a motion to dismiss.
• Notice matters. For reputation management, that changes the value of good documentation.
• Screenshots, dates, emails and correction requests are no longer merely administrative records.

A recent federal court ruling involving Google AI Overviews got our attention.

Not because Google has been found liable for defamation. It hasn’t.

Google was told the information was wrong. More than once. Yet similar false information continued to appear.

What happened

James “Jimmy” Keene is an author and television producer whose story became the Apple TV series Black Bird.

Google’s AI Overviews allegedly gave searchers very different information about him.

• One Overview said Keene was serving life in prison without parole.
• Another said he was serving a life sentence for murdering three women.

Not true.

The actual record:

• Keene had been convicted of conspiracy to distribute drugs in 1997.
• He later worked as a prison informant for the FBI.
• His conviction was eventually expunged.

According to the court record, Keene contacted Google three times about the false AI Overviews — on May 27, 2025, on May 31, and again after similar information continued to appear. That history mattered.

On September 14, 2026, U.S. District Judge Thomas Durkin refused to dismiss two of Keene’s defamation claims. The judge found that Keene had made a plausible case that Google could have known the information was false or acted with reckless disregard for whether it was false.

This is still an early-stage ruling. Two other Overview claims were dismissed. Keene has not won his lawsuit. Google has not been found liable. Still, we think the decision is important.

The bigger lesson is about notice

The most useful lesson for reputation management professionals like us is not that artificial intelligence sometimes gets facts wrong. Everyone knows that.

The more important question is what happens after a company learns that its AI is publishing false information about someone.

Keene alleged that he repeatedly told Google about the problem.

The court said those allegations were enough, at this stage of the case, to support his claim of “actual malice.”

That legal term can be confusing.

It doesn’t mean Google disliked Keene or tried to hurt him.

In defamation law*, actual malice generally means publishing something while knowing it is false or acting with reckless disregard for whether it is true.

Because Keene was treated as a limited public figure in this case, he had to meet that higher standard.

His repeated notices to Google helped him get there.

The court specifically noted two possibilities:

• Perhaps Keene’s reports reached people at Google who knew the information was false but failed to correct it.
• Or maybe nobody responsible reviewed the legal removal requests and related correspondence.

At the motion-to-dismiss stage, the court found either possibility could potentially support Keene’s claim.

Correction requests are now evidence

A correction request isn’t just a correction request.

When an AI system publishes a serious false statement about a person or company, reporting the error should be treated as a formal process.

• Save the evidence.
• Take screenshots.
• Record the exact search query.
• Record the date and time.
• Save Google’s response.
• Keep confirmation numbers.
• Save follow-up emails.
• Then check the same search again later.
• If the false statement returns, document that too.

Now you have a timeline.

A year ago, that timeline might have been useful mainly for helping an ORM team track an unresolved problem. Today, it could become evidence in litigation.

The Keene ruling makes that possibility more concrete.

Escalate serious errors faster

Not every AI error deserves a lawyer. AI systems make plenty of harmless mistakes. But some errors are different. A false claim that someone committed murder is obviously serious. So is a false accusation of:

• fraud
• sexual misconduct
• professional malpractice
• theft or another crime

Companies should treat those situations differently from ordinary reputation problems.

Start with the platform’s normal correction or feedback process. Just make sure you document it.

If a serious false statement remains after reasonable attempts to correct it, consider escalating the matter through a formal written notice. In particularly damaging cases, legal counsel may be appropriate.

The important point is to create the record while events are happening. It’s a pain to try to reconstruct it six months later, and by that time the information has most likely changed or gone stale.

What this could mean for AI platforms

This case may also change how AI companies handle correction requests.

The problem is not simply that an AI system can make a mistake.

No large generative AI system is going to achieve perfect accuracy.

The greater risk may come when a platform receives clear evidence that something is false and then continues publishing the same claim.

That creates a different situation.

Platforms may therefore have a strong incentive to improve the systems behind their correction forms:

• Serious reputation complaints may need faster human review.
• Repeat complaints may need automatic escalation.
• Corrections may need to propagate more reliably across the systems generating AI answers.

What we would tell a client today

If an AI Overview contains a materially false statement about you or your company, don’t treat it as a strange Google glitch. Treat it as a reputation issue that needs a documented response.

• Report it.
• Save everything.
• Follow up.
• Check whether it returns.
• Escalate serious unresolved errors rather than allowing them to sit for months.

The September 2026 ruling does not mean every incorrect AI Overview is defamatory. Nor does it mean everyone who complains to Google suddenly has a winning lawsuit.

What it does show is that notice matters.

For reputation management, that changes the value of good documentation.

In the right case, they may be the evidence that matters most.

• We at Reputation X are not attorneys.

Sources

• Keene v. Google LLC, No. 25 C 11431 (N.D. Ill. Sept. 14, 2026) — opinion (PDF) · Justia
• Volokh, E. “Federal Court Allows Libel-by-AI Claim to Go Forward.” Reason, Sept. 14, 2026.