AI Is Writing Your Bio, and It Is Getting It Wrong

Try something before you read any further. Open ChatGPT, or Gemini, or Perplexity, and ask it to tell you about yourself by name. Then ask it a few follow-ups, where you went to school, what your company does, what you are known for. If you are a heavily covered public figure, you will probably get something reasonable. If you are like most of the executives, founders, and candidates I work with, you are going to see something unsettling. The machine will answer with complete confidence, and a meaningful chunk of it will be wrong.

I do this exercise with clients constantly, and the reaction is always the same. They expected the AI to say “I don’t have much information.” Instead it produced a smooth, authoritative little biography, mixed a few real facts with a confused detail and an outright invention, and presented the whole thing in the same even, trustworthy tone. That is not a glitch. That is how these systems behave, and it is a reputation problem with a brand-new address.

Why the machine makes things up about you

Here is the part people find hard to believe until they see it. AI language models are not built to know whether what they are saying is true. They are built to produce text that reads as plausible. When the model has plenty of accurate material about a subject, plausible and true tend to line up. When it does not, the model does not stop. It generates something that sounds right anyway.

Researchers have a blunt name for this. A pair of academics at the University of Exeter describe it as “AI gossip,” noting that these systems routinely fabricate biographies and other details, producing fluent text that looks truth-apt without any real concern for whether it is true (arXiv). Others have argued the tendency to hallucinate is effectively built into how the technology works and will not simply be patched away (Scientific American).

And there is a cruel twist for exactly the people I tend to work with. The models are most likely to invent precisely the kind of specific, granular detail that makes a claim feel authoritative. A confident date, a named employer, a tidy career arc. So the thinner your public record, the more the AI has to invent, and the more convincing its inventions sound. The people with the most to lose from a wrong answer are the ones most likely to get one.

Who this hits hardest

This is not only a problem for the famous. It is the opposite. The better documented you are, the safer you tend to be, because the model has real material to work from. The danger lives with the people who have accomplished a great deal but have not been heavily written about.

That describes a lot of serious people. The founder whose company is well known but whose own story barely exists online. The senior executive who has spent a career being effective rather than visible. The down-ballot or first-time candidate who is new to public life and has almost no footprint for an AI to read. For all of them, the AI is not summarizing a record. It is improvising one, and then narrating it with the same authority it would use for a head of state.

The reason this now matters so much is that the AI answer increasingly comes first. Before a buyer takes the meeting, before an investor returns the call, before a reporter writes the profile, before a voter makes up their mind, more and more of them are quietly asking an assistant who you are. They arrive already holding whatever the machine told them, accurate or not. You are walking into rooms where the introduction has already been made, by a system that may have gotten you wrong.

You cannot argue with it, so do not try

The instinct, once people see a bad answer, is to want to correct the machine directly. Tell it it is wrong. Report it. Argue.

That is wasted energy, and it misunderstands the problem. The model is not consulting a profile of you that you can edit. It is generating from patterns across everything it has absorbed about you and people like you. You do not fix the answer by talking to the assistant. You fix it by changing what the assistant has to draw from.

That distinction is the whole game, and it is genuinely good news, because it means the lever is in your hands. The work looks like this.

It starts with seeing exactly what the assistants say today, captured word for word, including the confident errors. Then it is finding the gaps the model is filling, because those blanks point precisely at where your public record is thin, missing, or contradictory. From there it is publishing a clear, authoritative, well-structured account of who you actually are, the kind of source a machine can read cleanly and trust, and then reconciling the conflicting details scattered across your profiles and any third-party pages so every source tells one coherent story. And because the models keep changing, it is rechecking and maintaining rather than fixing once and walking away.

If you want the broader context on how AI now narrates people and brands, our piece on what ChatGPT says about you goes deeper, and the mechanics of being read and cited correctly are covered in what answer engine optimization is. For public figures and candidates specifically, this connects directly to political reputation management.

The quiet stakes

What makes this insidious is that it is invisible to you. A bad search result you can at least find and look at. A wrong AI answer is delivered privately, to one person, in a conversation you will never see, and then it is gone, having quietly shaped how that person regards you. Multiply that by every buyer, investor, reporter, and voter who looks you up this way, and you have a reputation being written, one confident hallucination at a time, entirely outside your view.

The fix is not glamorous and it is not instant. It is the patient work of making the truth about you clear, consistent, and easy for a machine to find, so that when the AI is asked, it has something real to say. That is the work we do, and for anyone whose record has not kept pace with their accomplishments, it is worth starting before the next important person quietly asks a chatbot who you are.

Sources

• arXiv — AI Gossip (Krueger and Osler, University of Exeter)
• Scientific American — AI Chatbots Will Never Stop Hallucinating
• arXiv — Caveats and Mitigations for Individual Users of AI Chatbots