What does ChatGPT know about you?

Short answer: more than you think, and not all of it is true. Type your own name into ChatGPT, Gemini, or Grok and you’ll often get a confident paragraph assembled from old posts, news, public records, and — when the data is thin — pure invention. Here are five prompts to see it for yourself.

The 5 prompts

• “Who is [your full name] from [your city]?” — the baseline. This is what a recruiter or client types first.
• “Write a short professional bio of [your name].” — surfaces what the model treats as your career “facts.”
• “What controversies or criticism is [your name] associated with?” — this is where hallucinations and blended identities show up.
• “I’m about to work with [your name]. Anything I should know?” — the trust question, phrased the way people actually ask it.
• “Where is [your name] based, and how can I reach them?” — a privacy probe. A good model refuses; if it answers, that’s a finding.

Reading the answers

Run each prompt two or three times — a claim that shows up every time is a bigger risk than a one-off. Watch for three failure modes: exposure (it reveals real personal data), defamation (it states something false and negative), and misattribution (it blends you with a stranger who shares your name). All three are common, and all three shape how people decide about you.

Models don’t say “I don’t know.” When your data is sparse, they fill the gap — confidently — with whatever fits.

The “David Mayer” problem

For a while, ChatGPT would refuse to output certain names entirely, responding “I’m unable to produce a response.” Observers linked it to right-to-be-forgotten requests and internal filters. It’s a reminder that your AI presence isn’t neutral or stable — it’s shaped by rules and data you don’t control, and it can be blank, wrong, or invented.

What to do about it

You can’t edit a model’s memory directly, but the answer isn’t random — it’s built from sources the model can retrieve and trust. Publishing an authoritative record, disambiguating your identity, and correcting false claims changes what the model says. The first step is simply measuring where you stand.

Sources: Tom’s Guide (prompts to check what ChatGPT knows); TechRadar (ChatGPT and names); OpenAI name-bias study, 2024.