How to check if ChatGPT recommends your brand (and why one prompt isn't enough)

Quick summary

• Ask ChatGPT what a buyer who doesn't know you yet would ask, not "what is my brand?". Brand questions only tell you whether ChatGPT knows you exist, not whether it recommends you.
• Use a clean session: a temporary chat or a logged-out window, with memory and custom instructions off. Otherwise ChatGPT may factor in what you've already told it.
• Ask each question several times. Answers change from run to run; with a single query you can't tell a real result from luck.
• Record mention, citation and competitors. If you're missing, first check that your robots.txt doesn't block OAI-SearchBot, the crawler ChatGPT uses for search.

Why asking about your own brand doesn't tell you much

Almost everyone starts the same way: open ChatGPT, type "what do you know about Acme?", and relax if the answer is right. But that question measures something else. A buyer who hasn't heard of you doesn't type your name; they type their problem:

• "Which Webflow agency would you recommend for a SaaS startup in Austin?"
• "Best booking software for a small dental practice"
• "Cheaper alternatives to [competitor] for a five-person team"

Those are the questions that decide whether you make the shortlist. Write ten to fifteen of them, with commercial intent and specific enough that the answer is a recommendation. Keep each in one language: the English and Spanish versions of the same question are two different questions with different answers.

How to set up a session that won't fool you

ChatGPT can personalize answers for logged-in users: memory, custom instructions, past conversations. If you've been talking to it about your company for weeks, it's no surprise it mentions you. That doesn't mean it mentions you to a stranger.

To reduce that bias:

1. Use a temporary chat or a private window while logged out.
2. Turn off memory and custom instructions if you test with your own account.
3. Don't stack questions in one conversation: one question, one new chat.
4. If a specific country matters, say so ("in the UK", "in the US"): location changes the answer.

None of this removes variability, but it stops you from measuring your own footprint.

Repeat: one answer is an anecdote

Ask the same question three times in three new chats and compare. Brands, order and sources usually shift. That's why "your position in ChatGPT" isn't a meaningful number, and why a single screenshot proves nothing.

A reasonable routine:

• Three runs per question, every week.
• Pool several weeks before deciding anything.
• Express the result as a share: "mentioned in 5 of 9 answers".

With samples this small, that share has a wide margin of error. If you want to quantify it, the Wilson interval works well with little data; we explain it step by step in how to measure your brand's visibility in ChatGPT.

What to record for each answer

One row per answer:

Question	Run	Mentioned?	Cited (link)?	Competitors shown	Cited URLs
best Webflow agency for SaaS	1	yes	no	Studio A, Studio B	roundup-x.com/…, studio-a.com/…

Three details worth not skipping:

• Mentions and citations are different. ChatGPT can name you without linking you (because it cites a third-party roundup) or link you without naming you (because your page is the source for a fact).
• The cited URLs are the most useful data you'll collect. They tell you where ChatGPT gets its answer for that question. If it cites three comparisons and you're in none, you know where to look.
• Watch the framing. "Acme is an option" and "Acme, although pricey" aren't the same. If ChatGPT repeats something wrong about you (an old price, a discontinued service), find the source it comes from.

If you're missing: what to check first

1. Can ChatGPT crawl you?

OpenAI documents three separate agents, and they're easy to confuse:

• OAI-SearchBot is used to surface websites in ChatGPT's search features. OpenAI says sites that opt out won't be shown in ChatGPT search answers (though they can still appear as navigational links) and recommends allowing it.
• GPTBot crawls content that may be used to train OpenAI's foundation models. Blocking it is a legitimate choice and independent of search.
• ChatGPT-User visits a page when a user asks ChatGPT to. OpenAI says it isn't used to decide what appears in search.

OpenAI also notes that robots.txt changes can take about 24 hours to take effect in search. You can check your file with the AI crawler checker. And watch your CDN or firewall: sometimes robots.txt allows OAI-SearchBot, but bot protection blocks it before it gets there.

2. Does your site answer the question?

If the question is "best booking software for dental practices" and your site only says "management solutions for healthcare", there's nothing to cite. Pages that get cited tend to answer a specific situation directly, with verifiable facts.

3. What do others say about you?

For many "best X" questions, ChatGPT leans on third-party comparisons and roundups. If your competitors are in them and you're not, the gap isn't on your site. We cover it in why AI cites your competitors instead of you.

When a tool is worth it

By hand, fifteen questions times three runs is 45 queries a week for ChatGPT alone. Add Gemini, Perplexity and Google and it multiplies. That's when automating pays off.

Every week, Citable asks your site's buyer questions in the AI engines on your plan, three times per engine, in sessions with no history. ChatGPT is read the way a user sees it. It separates mentions from citations, reports the share with its confidence interval and records which competitors and third-party pages appear instead of you. To try it first, the free report measures up to five questions in ChatGPT, Gemini and Google AI Overviews, no card required. Details are in the methodology.

What no tool can measure, ours included: each logged-in user's personalized answers. Those aren't visible from outside.

FAQ

Does ChatGPT give everyone the same answer?

No. Answers vary between runs and can be personalized for logged-in users, for example through memory. That's why you measure in clean sessions and repeat each question several times.

Will blocking GPTBot remove me from ChatGPT?

Not from search, according to OpenAI's documentation: GPTBot is used for training and OAI-SearchBot for ChatGPT search, and each is controlled separately in robots.txt. If you want to appear in search answers, don't block OAI-SearchBot.

Can I pay to be recommended by ChatGPT?

OpenAI doesn't document any way to pay to appear in its search answers. Be wary of anyone promising it. What you can do is stay crawlable, publish clear answers to your buyers' questions and be present on the sources ChatGPT already cites.

How many times should I repeat each question?

Three times per week per engine is a good starting point. More runs narrow the margin of error, and so does pooling several weeks. What matters is not deciding anything from a single answer.

Conclusion

Finding out whether ChatGPT recommends you isn't opening a chat and asking about your brand. It's asking a stranger's questions, in clean sessions, several times, and writing down who shows up and which sources get cited. If you're missing, start with what you can verify: OAI-SearchBot can get in, your site answers the question, and you're present on the third-party pages ChatGPT already uses.