How to Check if ChatGPT Mentions Your Brand

When someone asks ChatGPT "what's the best tool for X", the names in the answer become a shortlist. So the first question is simple: does ChatGPT name you? The second is why not, if it doesn't.

This tutorial is the manual check, done properly: how to set up the test so your own history doesn't skew it, which questions to ask, how to read the answers, and how often to repeat it. It uses MentionAgent's own AI visibility scans to show why a single ask can mislead you.

Set up a clean test

• Use a temporary chat. OpenAI's memory documentation recommends a temporary chat "when you do not want the conversation to use or create memories". If you have talked to ChatGPT about your own company before, a normal chat can carry that into the answer.
• Start a new chat for each question. An earlier question in the same chat shapes the next answer.
• Note whether the answer has sources. ChatGPT "may search the web automatically when your question would benefit from current information", per OpenAI's help center, and you can also choose Search yourself. An answer with source links came from a search; one without came from the model's training. Run each question both ways: once as a normal question, once with Search selected.
• Ask each question at least three times. Answers vary between asks; the rate tells you more than any one answer.

Ten prompts to run

Substitute your category and your main competitor where shown.

1. "What are the best [category] tools?" The plainest buyer question. Note your position if you're named.
2. "Recommend a [category] for [specific use case]." Tests whether ChatGPT ties you to a particular job, not just the broad category.
3. "What's an alternative to [main competitor]?" Buyers who already know a competitor ask this. Being named here puts you in a direct comparison.
4. "Compare [your brand] and [competitor]." Tests whether ChatGPT can describe you accurately when you're named in the question. Read the description, not just the verdict.
5. "Who makes [your product type]?" A category-leader question. If the answer is a list that doesn't include you, the association isn't there yet.
6. "I need a [category] that does [specific feature]. What do you recommend?" Feature questions can surface brands the generic list misses.
7. "What do people say about [your brand]?" Shows what ChatGPT finds in reviews and discussions about you, if anything.
8. "Tell me about [your brand]." Tests whether ChatGPT can describe you at all, and how accurately. Try it with and without Search.
9. "What's the cheapest [category] for [type of buyer]?" Price claims are often out of date, but the names are still worth noting.
10. "How do I [solve the problem your product solves]?" A problem question. In our scans these were the hardest for small brands; more on that below.

Our list of 22 prompts to test AI brand visibility has more, grouped by question type.

How to read the result

Every answer falls into one of three outcomes:

• Named and described correctly. You're in the list, with the right category and features. Keep doing what got you there and keep measuring.
• Not named, or named only when the question includes you. You have a presence problem. The pages ChatGPT reads, or the text it learned from, don't connect you with your category strongly enough.
• Named with wrong facts. A competitor's feature, an old price, the wrong category. This can be worse than not being named, because the buyer gets a wrong first impression. The fix is accurate descriptions of your product on the pages that cover your category, so there is correct text to draw on.

Check the sources it cites

On answers with source links, click every one and keep the list. It is the most direct evidence you will get of which pages ChatGPT reads for your category.

In MentionAgent's scans those were mostly other people's articles: on "best [category]" and "[competitor] alternatives" questions, two-thirds of ChatGPT's searched answers cited at least one list, review or comparison article. Each one that leaves you out is a page worth pitching. Our guide to listicle outreach covers how.

If ChatGPT cites other sites but never yours, check that OpenAI's search crawler can reach you. OpenAI's crawler documentation says sites that opt out of OAI-SearchBot "will not be shown in ChatGPT search answers".

Common reasons ChatGPT doesn't name you

• It answered from memory. No sources means no search, and small brands rarely appear in those answers. In our scans ChatGPT's model skipped the search on most "how do I" questions.
• The articles it read leave you out. The most common and most fixable cause.
• Your competitors are named everywhere and you aren't. More independent pages name them alongside your category than name you.
• Its search crawler is blocked by your robots.txt or a CDN bot rule.
• Your name is ambiguous or new. A common-word name competes with its everyday meaning, and a brand launched after a model's training won't be in answers given from memory.

Each cause, with the numbers behind it and a test for which one is yours, is covered in why ChatGPT doesn't recommend your business.

Checking at scale

Ten questions, asked three times each, two ways, is 60 answers per engine. That is fine once. Done every month, across ChatGPT, Gemini and Claude, it becomes a spreadsheet job. You have four options:

• A manual log. For each round, record the question, whether you were named, your position, the description and the sources. The free ChatGPT rank tracker keeps the named and position columns for you: paste the answer and it logs both.
• MentionAgent's AI visibility scan. Included with a subscription: it asks a panel of about 15 buyer questions to ChatGPT, Gemini and Claude twice a month and tracks whether you were named and which sources were cited.
• The free AI Mention Checker. It does not ask ChatGPT anything. It checks your brand's presence on Wikipedia, Hacker News and GitHub and your domain's PageRank, a quick look at a few public signals before you test.
• A dedicated AI visibility platform. Aiso uses insights from what it describes as millions of real, consent-based AI-human conversations to understand what people ask AI assistants and which brands get recommended. It pairs that analytics with end-to-end services to help brands improve their presence in AI search.

How often should you check

Monthly while you're actively working on it, quarterly otherwise. Searched answers can change as soon as the pages ChatGPT finds for a question change, so active outreach can show up quickly. Answers from training data only change when OpenAI releases a new model.

Keep the log. Over a few rounds the patterns become clear: which sources keep appearing, which questions you're gaining or losing, and whether the description is settling in your favour.

How the other engines compare in our data

Checking ChatGPT alone gives you one engine's view. Across the same buyer questions, the engines behaved very differently:

ChatGPT's model was the only one that regularly answered without searching. That is why testing it both ways matters more than for the other two. For engine-specific guides, see Gemini, Claude and Perplexity, and the ChatGPT playbook for what to do once you've found the gap.

Method

• Source. MentionAgent's AI visibility scans for 41 customer brands, 8 July to 5 October 2026: about 15 buyer questions per brand, none containing the brand's name, asked to each engine twice a month.
• ChatGPT in our data is the API, not the app. The ChatGPT app uses other models and can draw on a user's history, so its answers will differ. The manual test on this page is how you check the app itself.
• "Named". The brand's name appears in the answer text. "No sources" means the answer cited no URLs.

Frequently asked questions

Related guides