What does AI say about me?

How to find out, in five minutes

1. Ask each assistant the plain question. Type "Who is [your full name]?" into ChatGPT, Claude, Gemini and Perplexity. Use a private window or a fresh chat, so your own history does not shape the answer.
2. Turn web search on. Most people who look you up use the app's default, which searches the web. Answers without search come only from training data and are often older.
3. Note the sources. Perplexity, ChatGPT search and Gemini cite pages. Those pages are what you can influence.
4. Try a namesake test. Add your employer or city: "Who is [name], [title] at [company]?" If the answer changes, the assistant is unsure which person you are.

The free SelfBadge AI check does all of this at once: it asks the four engines the same question with web search on and shows their answers side by side, word for word, with the sources each one cited.

The five ways AI gets people wrong

• Namesakes merged: two people with the same name become one, with the other person's job, age or controversy attached to you.
• Stale roles: the assistant repeats a job you left years ago, because older pages outnumber newer ones.
• Missing entirely: "I don't have information about this person", even though you have a LinkedIn profile, because it is behind a sign-in wall.
• Wrong links: an account that is not yours is presented as yours.
• Invented details: without a clear source, a model fills gaps with plausible-sounding facts.

Why it happens

An assistant answering about a person has to decide which pages are about the same individual, then summarize them. It has little to go on: a name, a few job titles, perhaps a photo caption. Pages that state facts clearly, link the person's accounts together and carry structured data (schema.org markup) are far easier to match and quote than a loose collection of mentions. Read more in how AI assistants identify people.

What changes the answer

1. One clear, open source about you. A page that states who you are in plain sentences, readable without a sign-in.
2. Proof that your accounts belong to you. Links that go both ways (your site links to your profile, your profile links to your site) and verified accounts.
3. Structured data. schema.org Person markup with a stable ID, your employer, your role and sameAs links to your real accounts.
4. Consistency. The same name, title and employer everywhere you control.
5. A way to tell you apart. One line that separates you from people with the same name: field, city, employer.

A verified SelfBadge profile covers the first four in one place, and has a "to tell them apart" line for the fifth. It is free, and it is published in the formats assistants read: JSON-LD in the page, a Markdown version and a public API.

How long until AI notices?

Assistants with web search can pick up a new page within days of it being crawled. SelfBadge notifies search engines through IndexNow when you publish or change your profile. Answers that come only from training data change when the next model version is trained, which can take months. Check again after a couple of weeks.