How to check if ChatGPT mentions your brand
If you want to know whether ChatGPT brings up your product when people ask about your category, the honest answer is that you have to ask it, carefully and more than once. There is no official report from OpenAI that tells a site owner how often their brand appears in answers. What you can do is run a small, repeatable test by hand and write down what you see.
This guide covers how to set that test up so the results mean something, and where its limits are.
Step 1: Decide what you're actually asking
Write down 5 to 10 prompts that a real customer might type. Mix three kinds:
- Category questions: "What are good tools for scheduling podcast guests?"
- Problem questions: "How do I stop double-bookings when I run a small salon?"
- Comparison questions: "What are alternatives to [the biggest competitor]?"
Avoid putting your own brand name in most of them. "Is [your brand] good?" tells you what ChatGPT says about you when asked directly, which is useful, but it doesn't tell you whether you come up on your own. Keep one or two direct-name prompts as a separate group so you can check that the description of your product is accurate.
Save the exact wording. You'll reuse it word for word every time you test.
Step 2: Make sure the answer uses web search
ChatGPT can answer from what its model already knows, or it can search the web and cite pages. OpenAI says ChatGPT may search automatically when a question benefits from current information, and you can also start a search yourself: open the tools menu and choose Search, or type / in the message box and select Search (OpenAI Help: Searching the web with ChatGPT).
For a brand check, searched answers are the more useful ones, because they show the sources ChatGPT relied on. Pick one approach and stick to it. If some runs search and others don't, you're comparing two different things.
Step 3: Reduce personalization
Your own account can tilt the results. OpenAI documents that if memory is on, ChatGPT may use saved memories when it rewrites your question into search queries, and that it uses an approximate location based on your IP address for local results (OpenAI Help: Searching the web with ChatGPT). If you've talked to ChatGPT about your own company, that history is the last thing you want shaping the test.
Two ways to reduce this:
- Use a temporary chat set to Unpersonalized. OpenAI says you choose Personalized or Unpersonalized before sending the first message, and that Unpersonalized doesn't use memory, custom instructions, or plugins (OpenAI Help: Temporary chat). You can't change that choice once the conversation starts, so set it first.
- Test while logged out. OpenAI says people who aren't signed in can also use web search (OpenAI Help: Searching the web with ChatGPT).
Neither option removes location. If location matters for your business, decide which city you're testing for and put it in the prompt, which is what OpenAI suggests when local results show the wrong area.
Step 4: Run each prompt several times
Don't judge from a single answer. In a January 2026 study, SparkToro and Gumshoe.ai had 600 volunteers run 12 recommendation prompts through ChatGPT, Claude, and Google's AI results a combined 2,961 times. They reported a less than 1 in 100 chance that ChatGPT or Google's AI would return the same list of brands in any two responses, and that the length of the list varied too (SparkToro, 27 Jan 2026). It isn't peer-reviewed, and one co-author works at an AI-tracking startup, which the post discloses. But the practical point holds for anyone testing by hand: one run is an anecdote.
A workable habit for a small team is to run each prompt at least three to five times per session, each in a fresh temporary chat. That won't give you statistics, but it shows whether you appear sometimes, often, or never. If you need firmer numbers, the same SparkToro post argues for far more runs per prompt.
Step 5: Record the same things every time
A simple spreadsheet is enough. One row per run:
| Column | What to write |
|---|---|
| Date and time | When you ran it |
| Prompt | The exact saved wording |
| Setup | Temporary unpersonalized, or logged out; search on or off |
| Mentioned? | Yes or no |
| Position and wording | Where you appeared and how you were described |
| Others named | Every competitor in the answer |
| Cited sources | The URLs ChatGPT linked |
The last two columns matter most over time. They tell you who is being recommended instead of you and which pages the answer leaned on.
Step 6: Open the cited sources
Answers that use search can include citations, and OpenAI says you can select one to open it, or select Sources to see the cited pages and other relevant links (OpenAI Help: Searching the web with ChatGPT).
Open them. You're looking for patterns:
- Are the cited pages review sites, comparison articles, forums, or the competitors' own sites?
- Is your product listed on those pages at all?
- When you are mentioned, is the description current and correct?
OpenAI also says search results and citations can be incomplete, outdated, or incorrect, and recommends opening the source to check it supports the answer (same page). So if ChatGPT describes your pricing wrongly, the cited page is the place to start looking for why.
Step 7: Repeat on a schedule
Run the same prompts on the same setup every few weeks, or after a launch or a big change to your site. Compare the "mentioned" rate and the list of cited sources over time, not single answers.
What this test can't tell you
Be clear about the limits before you share the results with anyone:
- It's a sample, not a measurement. Real people phrase questions in many different ways. Your 10 prompts are a narrow window.
- Answers change without warning. Models, search partners, and ranking change. OpenAI says ChatGPT ranks search results using multiple factors and that placement is not guaranteed (OpenAI Help: Searching the web with ChatGPT).
- It doesn't explain why. A "no" tells you that you weren't named, not what caused it.
- It's ChatGPT only. Other assistants use their own crawlers and sources. Anthropic and Perplexity each document separate bots for search and user-requested fetches (Anthropic, Perplexity), so results in one tool don't carry over to another.
One thing to rule out first
If your site never appears and is never cited, check whether ChatGPT's search can reach it at all. OpenAI says sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers, though they can still appear as navigational links, and it recommends allowing OAI-SearchBot and its published IP ranges (OpenAI crawler documentation). A robots.txt rule or a firewall setting can quietly rule you out before content quality comes into it.
Checking the page side: whether you're mentioned is something you have to test by hand, as above. Whether your pages can be fetched and understood is a separate question, and that one you can check directly. CiteLens is a free, no-sign-up audit of a public page: robots.txt rules for documented AI crawlers, llms.txt, sitemaps, structured data, title and H1, static text, FAQ coverage, and Open Graph. It doesn't run prompts or tell you whether any assistant mentions you.
Disclosure: Born Bots builds CiteLens.
Sources
- OpenAI Help Center, "Searching the web with ChatGPT": https://help.openai.com/en/articles/9237897-chatgpt-search (opened 17 Nov 2026)
- OpenAI Help Center, "Temporary chat in ChatGPT": https://help.openai.com/en/articles/8914046-temporary-chat-faq (opened 17 Nov 2026)
- OpenAI, "Overview of OpenAI Crawlers": https://platform.openai.com/docs/bots (opened 17 Nov 2026)
- SparkToro, "NEW Research: AIs are highly inconsistent when recommending brands or products", 27 Jan 2026: https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/ (opened 17 Nov 2026)
- Anthropic (Claude Help Center), "Does Anthropic crawl data from the web, and how can site owners block the crawler?": https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler (opened 17 Nov 2026)
- Perplexity, "Perplexity Crawlers": https://docs.perplexity.ai/guides/bots (opened 17 Nov 2026)
Last reviewed: 17 Nov 2026.