AI Hallucinations: What to Do When AI Says Something False About You

An AI hallucination is when an AI tool states something false as if it were fact, such as an invented lawsuit, a wrong job title or a quote nobody said. If an AI tool says something false about you, document it, report it through the product’s feedback or privacy channels, check whether any web source is feeding the claim, and strengthen accurate sources about you. If the claim is serious and keeps appearing, talk to a lawyer, because the law on AI-generated falsehoods is still developing.

This guide explains why hallucinations happen and what to do about one that affects your name or business. For the broader work of shaping what AI tools say about you, see our guide to managing your reputation in AI search.

What an AI hallucination is

AI chat assistants and AI search features write answers by predicting likely words based on patterns learned from huge amounts of text, sometimes combined with pages they fetch from the web. They don’t look facts up in a verified database the way a phone book works. Most of the time the prediction lines up with reality. Sometimes it doesn’t, and the tool produces a fluent, confident sentence that isn’t true.

The term “hallucination” covers a range of errors:

• Invented facts. A lawsuit, arrest, award or bankruptcy that never happened.
• Merged identities. Details from two people or businesses with similar names blended into one answer.
• Fabricated sources. A citation to an article that doesn’t exist, or a real link that doesn’t say what the answer claims.
• Distorted summaries. A real source misread, so “was cleared of the charge” becomes “was charged.”
• Invented quotes. Words attributed to you that you never said or wrote.

Why AI tools make things up about people and businesses

A few conditions make hallucinations about you more likely:

1. Thin information. When there isn’t much written about you, the model fills gaps with plausible-sounding detail drawn from similar people or businesses.
2. A common or shared name. If someone with your name has a lawsuit or a news story, the model can attach it to you. Our guide for people with the same name as someone with a bad reputation covers how to separate yourself in search.
3. Outdated training data. Models learn from the web as it was at training time. A story that was later corrected can live on in older answers.
4. Leading questions. A question like “Why was [business] sued?” can push a model to supply a reason, even when no lawsuit exists.
5. Retrieval mistakes. AI search tools can fetch the wrong page, or pull a sentence out of context from the right one.

None of this is deliberate. But the effect on your reputation can be the same as a false article: a customer, employer or journalist reads it and believes it.

Hallucination, bad source or accurate but negative?

Before you act, work out which problem you have. The fix is different for each.

What you see	What it usually is	Main fix
A claim with no source, or a source that doesn’t say it	Hallucination	Report it, document it, strengthen accurate sources
A claim that matches a real page, but the page is wrong	Bad source	Correct or remove the source page
A claim that matches a real page that’s out of date	Outdated source	Ask for an update, then refresh the search result
A claim that’s true but unflattering	Accurate negative information	Context, new material, operational fixes

To tell them apart, open every source the answer cites and search the claim yourself on regular search engines. If nothing on the web says it, you are almost certainly dealing with a hallucination.

What to do if an AI tool says something false about you

1. Document it properly

Evidence matters if you later need a correction, a privacy request or legal advice.

• Take full screenshots showing the whole answer, the question and the tool’s name.
• Record the date, time, the exact prompt, and whether you were signed in.
• Copy any cited links, and note whether each one actually supports the claim.
• If the tool offers a share link for the conversation, save it.
• Try again in a clean session and with different wording. Note how often the false claim repeats. A claim that appears every time is more serious than a one-off.
• If someone else showed you the answer, ask them for their screenshot and the question they asked.

2. Report it through the product’s official channels

Every major AI tool has a way to flag a bad answer, usually a thumbs-down or report option on the answer itself. Use it, and write one or two plain sentences: what the answer said, why it’s false and what’s true. Some providers also accept privacy requests about inaccurate personal data. OpenAI, for example, accepts requests through its privacy portal to correct or remove personal data in ChatGPT’s answers, and reviews them case by case. Our guides to ChatGPT search, Google AI Overviews and Perplexity AI cover each product’s routes.

If you live in the EU or UK, data protection law gives you rights over inaccurate personal data, including a right to rectification. How those rights apply to AI models is still being worked out by regulators, but it is a legitimate basis for a request.

Feedback and requests don’t come with a promised outcome or timeline. Keep a log of what you sent and when.

3. Check and fix the sources around the claim

Even a pure hallucination often has a trigger: a same-name news story, an ambiguous directory entry or an old page that mentions you near a lawsuit. Find it and fix what you can.

• Correct your own pages and profiles so your role, business name, location and history are stated plainly.
• Ask third-party sites to fix wrong or ambiguous details, with evidence.
• If the confusion is with someone else, say so plainly on your own site: “Not affiliated with…” lines are dull but effective.
• Where a web page has changed, use Google’s Refresh Outdated Content tool so the search result updates. Our guide on fixing wrong information on Google covers the other correction routes.

4. Build accurate material to fill the gap

Hallucinations thrive on thin information. A clear bio or About page, consistent professional profiles, genuine reviews and independent coverage all give AI tools real facts to use instead of guesses. Structured data and entity work, covered in our guide to entity SEO, help tools recognize you as a distinct person or business.

5. Don’t spread the false claim yourself

A public post saying “ChatGPT says I was arrested for fraud, which is false” can put the words “arrested” and “fraud” next to your name on a page that ranks and gets picked up by the next AI answer. If you need to respond publicly, state the true facts without repeating the false claim.

When to talk to a lawyer

Most hallucinations are best handled through feedback and better sources. Consider legal advice when:

• The false claim is serious, such as a crime, fraud, professional misconduct or a health condition.
• It keeps appearing after you have reported it and fixed the obvious sources.
• You can show real harm, such as a lost contract, a withdrawn job offer or customers citing it.
• A person or website has copied the AI’s false claim and published it. A human publisher repeating a falsehood is a more familiar legal situation, and our guides on defamation of character and defamation vs opinion explain the general principles.

Bring your documentation log to the first meeting. A lawyer can tell you whether a formal request to the provider, a letter to a republisher or another step makes sense where you live. Avoid sending threatening letters yourself: they rarely help with AI providers and can backfire with human publishers.

A worked example

This is an illustrative scenario, not a real client. “Owen Castellano” runs a small wedding photography business. A couple tells him they almost didn’t book because an AI assistant said he “settled a lawsuit in 2022 over undelivered wedding photos.”

1. Owen asks the same question in several tools. One repeats the claim with no source. Another gives an accurate summary. He screenshots both, with dates and prompts.
2. He searches the claim on Google and Bing and finds nothing about him. He does find a local news story about a different photographer with a similar business name.
3. He reports the false answer through the tool’s feedback option and submits a privacy request explaining the claim is false and appears to be a mix-up.
4. He updates his website’s About page with his full business name, the year he started and a line clarifying he has no connection to the other studio, and makes his profiles on wedding directories match.
5. He keeps asking every couple for a review on the platforms he uses, which adds current, genuine material about his work.

Over the next couple of months, the false claim appears less often and then stops showing up in his regular checks. He keeps the documentation in case it returns. Had it continued, or had a blog repeated it, his next step would have been a conversation with a lawyer.

Common mistakes

• Arguing with the chatbot. Correcting it in your chat doesn’t change what it tells anyone else.
• Not saving evidence. AI answers change. If you don’t capture it the first time, you may never see it again, but a customer might.
• Assuming there’s a source. Hours spent hunting for the “article” behind a hallucination are better spent on reporting and accurate sources.
• Repeating the false claim publicly. It can create the very source that makes the claim stick.
• Trying to trick the models. Hidden text, fake reviews and planted posts break platform rules and the FTC’s 2024 rule on fake reviews, and create a real story where there wasn’t one.

If a hallucination is already costing you work and you want help with the documentation, source fixes and monitoring, our personal reputation management team can take that on, and the free audit is a good place to start.

Frequently asked questions

Can I sue an AI company for a hallucination about me?

Possibly, but it’s uncertain. A few cases have been brought, and courts have not settled how defamation law applies to AI-generated statements or who is responsible. Talk to a lawyer about your specific facts before taking any step.

Will reporting an AI hallucination make it go away?

Not necessarily. Feedback and privacy requests help providers fix errors, but there’s no promised outcome or timeline. Combining reports with accurate, consistent sources about you gives the best chance of the claim fading.

How can I tell if an AI claim about me came from a real source?

Open every citation the tool gives and check that the page actually says what the answer claims. Then search the claim on regular search engines. If nothing supports it, treat it as a hallucination.

Are AI hallucinations getting less common?

AI providers say they are working to reduce them, and tools that search the web and cite sources make errors easier to check. They still happen, especially for people and small businesses with little written about them, so regular checks remain worthwhile.