How to check what ChatGPT, Gemini and Perplexity say about you, and correct it

Updated 3 October 2026 · 8 min read

Clients now ask AI assistants about professionals before they pick up the phone. What do they say about you? They may name you correctly, leave you out, mix you up with someone else, or state an old phone number as fact. You cannot edit an AI assistant, but you can check what it says, find where a mistake comes from, and fix it at the source. This guide gives you a simple monthly method, twelve questions to ask, a way to record the answers and the routes for reporting errors.

The short answer

Ask each assistant the questions a client would ask, once a month, with web search switched on, and record the full answer, the date and the sources it shows. Answers vary, so one test proves little; a pattern over several months shows something.
The usual problems are being left out, being described with old or wrong details, and being confused with someone who has a similar name.
You cannot edit the assistant. You fix the sources it reads: your website, your Google listing, your profiles and directories, and any old page elsewhere. Then report serious errors through the assistant’s own feedback route.
OpenAI accepts correction requests about inaccurate information on individuals, but says it will not be able to correct every inaccuracy. Google and others have feedback and legal-report options.
If a false statement damages your reputation or concerns your personal data, data-protection law in some places gives you a right to correction, and a lawyer can advise on defamation. This guide is not legal advice.

Why check

Clients ask. A person choosing a dentist or a lawyer may ask an assistant “who should I see in my city?” and act on the answer.
Mistakes spread. An old address or a wrong job title repeated by an assistant reaches people you never meet.
You find out what to fix. The answers show which sources the assistants trust and which of your facts are missing or inconsistent.
Silence is information too. If you are never mentioned, that points to a visibility gap you can work on.

Set up a fair test

Use the assistants clients are most likely to use. For most people that means ChatGPT, Google (AI Overviews and Gemini) and Perplexity; add Microsoft Copilot if your clients use it.
Switch on web search or browsing where the tool offers it, because the answers about current professionals come from the live web.
Start a new conversation for each question, so earlier answers do not colour the next.
Ask in the way a client would, not in technical language. Add your city, as they would.
Note that answers differ by account, location, wording and day. If you are signed in, past conversations or saved memory can influence results, so also try a private window or a signed-out session.
Record each test: date, assistant, exact question, the answer and the sources shown.

Twelve questions to ask

Replace the words in brackets with your own. Ask them in your own language as well if your clients use it.

Who is [your name]?
Who is [your name], [your profession] in [your city]?
Who is a good [profession] in [city]?
Who is a good [profession] in [city] for [your specialty]?
How much does [a service you offer] cost in [city]?
What should I ask before hiring a [profession] in [city]?
Is [your name] a real [profession], and are they qualified?
What are the opening hours and address of [your practice or business name]?
How do I book an appointment with [your name]?
What do people say about [your name] or [your practice]?
Compare [your name] with [a nearby colleague].
Which [professions] in [city] speak [a language you speak]?

Questions 1, 2, 7, 8 and 9 test whether the facts about you are right. Questions 3, 4, 5, 6, 11 and 12 test whether you come up when a client does not know your name. Question 10 tests what is said about your reputation.

Record the answers

A simple table is enough. Keep it in a spreadsheet and add a row for each test.

A record of what assistants say (example rows, invented)
Date and assistant	Question	Mentioned and correct?	Sources shown	Action
3 Oct 2026, ChatGPT with search	Who is a good dentist in Porto for children?	Not mentioned	A city directory and two clinic sites	Add “paediatric dentistry” and Porto to my page and directory listing
3 Oct 2026, Perplexity	What are the opening hours of Clínica Costa?	Mentioned, but shows old hours	My old directory entry	Update the directory, then check my site and Google listing

How to read the results

What each result means and what to do
What you see	What it usually means	First step
You are not mentioned	The assistant did not find, trust or match your page for that question	Check your page title names your profession and city, your page is crawlable, and the facts are on a page, not just in images
You are mentioned with wrong details	It read an old or mistaken source	Look at the sources it shows, correct each one, and update your own page and listings
It mixes you up with someone else	Your name is shared and your details are not distinctive enough	Add your profession, city and specialty everywhere you appear, and link your profiles to each other
A competitor is named, not you	They have clearer or more widely confirmed information for that question	Compare their page and listings with yours, and fill the gaps honestly
It is right, but gives no link	It may have used its training data, or a source that is not your site	Make sure your page says the same thing in plain text, and keep your listings consistent

Fixing mistakes at the source

An assistant is only as good as the pages it reads. So the lasting fix is to correct those pages.

Your own website. Make sure it states your name, profession, city, address, phone number and hours in plain text, and in structured data. See structured data for professionals.
Your Google listing. Correct and complete it. See how to get on Google Maps.
Your profiles and directories. Update LinkedIn, your professional body’s listing and any directory.
Old pages elsewhere. If a site you do not control still shows an old address or a closed practice, ask its owner to update or remove it.
Wait. There is no fixed timetable for an assistant to re-read a corrected page; it can take days or weeks. For comparison, OpenAI says changes to its crawler settings take about 24 hours to be reflected in ChatGPT search.

Reporting errors to the assistant

Google AI Overviews: each overview has thumbs up and down icons. Choose thumbs down, then “Report a problem” and the category that fits. Google also has a form for reporting content for legal reasons, which covers Search and its AI products.
ChatGPT: use the thumbs-down on the answer to say it is wrong. For inaccurate information about you personally, OpenAI’s privacy policy describes a correction request through privacy.openai.com. OpenAI says it will not be able to correct every inaccuracy; if it cannot, it suggests asking for your personal information to be removed from its results.
Gemini and Perplexity: most assistants offer a feedback or report option on each answer. Use it and describe the exact error.

Expect imperfect results. A privacy group’s complaint to European regulators in 2024 argued that ChatGPT could not correct false information about people, and OpenAI’s own wording acknowledges the limit. That is why fixing the sources matters most.

If it is serious

If an assistant states something false that damages your reputation, or wrongly shows your personal data, you may have legal options. Data-protection law in some places, such as the EU and UK GDPR (right to rectification) and California’s privacy law (right to correct), gives you rights to have inaccurate personal data corrected. A false factual statement may also be defamation. Keep screenshots with dates, and speak to a lawyer. This is general information, not legal advice.

If you share your name with someone else

Always use your name with your profession and city: “Maria Santos, Cardiologist in Lisbon”.
Use the same short description of yourself on every profile and listing.
Link your profiles to each other and to your website, and list them in structured data using sameAs.
Include something distinctive and checkable: your registration number, your clinic, your specialty.
Ask for corrections on the wrong person’s page only if it truly mislabels you. Do not try to remove another real professional.

How often to check

Once a month is enough for most people, plus an extra check after you change your address, phone number, name or services. Do not read much into a single answer. Look for patterns: are you mentioned more often than three months ago, are the details right, are the sources you control being used?

Be sceptical of tools that sell a single “AI visibility score”. The answers vary too much for a precise number to mean much. A table you keep yourself, with real questions and real answers, tells you more.

How BeVisible helps

BeVisible keeps the facts on your page, in structured data and in a plain-text summary for AI generated from the same details, so they cannot disagree. When you change a phone number or hours, you change it once. You can build and preview free, and publish free for 7 days with no card needed; after that one yearly plan keeps a website online, and one plan covers one website. Start from a photo or type your details.

Frequently asked questions

How do I see what ChatGPT says about me?

Open a new chat with web search switched on and ask the questions a client would, such as who you are, who is a good professional in your field in your city, and what your practice charges. Record the answer and the sources shown, then repeat on Google and Perplexity.

Can I make ChatGPT correct wrong information about me?

You can ask. OpenAI’s privacy policy describes a correction request for inaccurate information about individuals, and the thumbs-down on an answer reports a mistake. OpenAI says it will not be able to correct every inaccuracy. The lasting fix is to correct the sources the assistant reads, mainly your website and listings.

Why does the AI give a different answer each time?

These systems generate answers freshly, and they differ by wording, location, account, saved memory and the sources found that day. That is why one test proves little, and why a monthly record is more useful.

Why am I not mentioned at all?

Common reasons are that the assistant could not read your page, your page does not clearly name your profession and city, your facts do not match across sources, or other professionals have clearer or more widely confirmed information. Work through the first steps in the table above.

How do I report a wrong Google AI Overview?

Use the thumbs-down icon under the overview, choose “Report a problem”, and pick the category that fits. For legal issues, Google has a web form for reporting content for legal reasons.

Do I have a legal right to have wrong information about me corrected?

In some places, yes. EU and UK data-protection law includes a right to have inaccurate personal data corrected, and California’s privacy law gives a right to correct. Defamation law may also apply. This depends on your country and the facts, so speak to a lawyer.