What does AI actually know about you?
Less than you fear, and less than you'd want. AI assistants know two kinds of things about a person: what you've told them in your own chats (private, per app, fixable in settings), and what the public record says about you, which is what they tell anyone who asks. The second kind is effectively your reputation now, because asking an assistant “who is this person?” has quietly become how people check out a founder, a consultant or a freelancer before a call.
The only way to know where you stand is to run the check yourself. It takes about fifteen minutes across the four assistants most people actually use.
The 15-minute AI audit
Open each assistant in a fresh session (logged out, or in a temporary mode where offered, so your own account data doesn't color the answer) and ask the same two questions: “Who is [your full name]?” and “What does [your full name] do professionally?” Then compare:
- ChatGPT: with browsing it leans on live web results; without it, on training data. “I don't have information about this person” is a blank slate, not an error.
- Google Gemini and AI Overviews: the closest thing to Google's index of you. Search your own name in Google too, and note whether an AI Overview appears above the results.
- Perplexity: always cites its sources, so it shows you exactly which pages currently define you. Its citation list is your to-do list.
- Microsoft Copilot: draws on Bing's index, which is often thinner for individuals. A blank here usually means Bing barely knows your pages exist.
How to read the four answers
Whatever comes back lands in one of four buckets, and each has a different fix:
- A blank (“I don't have information”): nothing corroborates you yet. Annoying, but the easiest starting point, because there's nothing wrong to displace.
- A mix-up: the assistant blends you with a namesake. The fix is making your identity the better-corroborated one, not fighting the namesake's.
- Outdated: it describes the job you left or the company you sold. The sources it reads are stale, so refresh the ones you control.
- Accurate but thin: it knows your name and one fact. That's a foundation, and everything you add from here compounds.
Why each assistant says something different
Because they read different sources at different times. ChatGPT's browsing and Copilot lean on Bing's index, Gemini on Google's, Perplexity on its own retrieval, and all of them on whatever their training data froze months ago. There is no single database of you; there are several, all incomplete, all updating on their own schedules.
That's also why the fix is the same for all of them: publish one consistent, machine-readable version of who you are, connect your accounts so they corroborate each other, and let every engine read the same story. The playbook is in how to control what ChatGPT says about you, and the ChatGPT-specific deep dive is in does ChatGPT know who you are.
Check it without doing it by hand
Search Presence runs this audit for you: it checks what AI currently says about you with a live, grounded search, shows you the answer word for word, and then builds the verified identity page that gives every assistant the same accurate story to read.