OpenAI warns that ChatGPT can sound convincing and still be wrong, and that its knowledge is not up to date. With search, ChatGPT, Perplexity and Google's AI features show the pages they used. That is where the work happens.
How to check what they say about you
Ask the questions a client would ask, in more than one assistant, and note the answer and the sources. Repeat monthly: answers change.
- "Who is [name]?", on its own, as a stranger would ask.
- "Who is [name], [profession] in [city]?", with the context that identifies you.
- "Who would you recommend as a [profession] in [city]?", to see whether you appear in a recommendation.
- Record the cited sources: those are the pages worth fixing or strengthening.
The free check on this site asks the first question with web search and shows the sources. For "Emanuel Luís" on 1 November 2026, the answer was that several people share the name and none of them was the author of this guide.
Why AI gets it wrong
Three errors are most common: mixing up people with the same name, repeating old information from pages nobody updated, and inventing plausible details when there is little to go on. All three shrink when a clear, current source says who you are.
Fix the sources first
Assistants that search cite pages that are indexed. Google says that to be eligible for its AI features a page must be indexed and able to show a snippet; there are no additional technical requirements or special files or markup, but appearing is not guaranteed. OpenAI recommends allowing its OAI-SearchBot crawler.
- A page about you, with name, profession, location and links to your official profiles.
- Consistent profiles: the same name and description everywhere.
- Old pages with wrong details updated or taken down at the source.
- Don't block search crawlers (Googlebot, OAI-SearchBot). Blocking training crawlers such as GPTBot does not stop ChatGPT search from citing you.
Correction and removal requests to AI companies
Under the GDPR, you can ask these companies to correct or stop using your personal data. Requests are not automatic: OpenAI says it assesses each one case by case, weighing free expression and the public interest; Anthropic says it may decline where there is a public interest in keeping the information.
- OpenAI (ChatGPT): in the Privacy Portal, "Remove personal data from ChatGPT responses", with or without an account, with chat links and proof of identity. It only affects ChatGPT answers, not websites or search engines.
- Google (AI Overviews and AI Mode): the feedback buttons on the answer itself. Searchers cannot turn the feature off.
- Perplexity: its data privacy form or an email to the privacy team, which says it replies within 30 days or asks for an extension of up to two more months. Every answer includes citations.
- Anthropic (Claude): a request through its privacy centre, with screenshots, the prompt used, your name and email.
If a company refuses without good reason, you can complain to your data protection authority.