How to check what ChatGPT says about your business

How do you check what ChatGPT says about your business?

Ask it directly, in a fresh chat, the way a buyer would. Open a new conversation with no history (a temporary chat or logged-out session is best) and type a real buying question, such as "what is the best [your category] for [your customer]". Do not ask "what do you know about [my company]", because buyers never phrase it that way and the answer tells you little.

Then read the answer the way a stranger would. Note whether your brand appears at all, where it sits in the list, what ChatGPT says about you, and which sources it cites. When search is on, the citations tell you which pages are doing the talking for you.

Repeat the same question two or three times in separate fresh chats. You will almost certainly get different answers, and that variation is part of the finding, not a flaw in your method.

Which questions should you actually test?

Test the questions buyers ask before they know your name. Category questions ("best accounting tool for small construction firms"), comparison questions ("X vs Y"), and problem questions ("how do I stop invoices going unpaid") are where customers are won or lost. Your own brand name matters less, because someone who already types it has already found you.

A practical starting set is five to ten questions: two or three category questions, one or two comparisons against competitors you actually lose deals to, and a couple of problem questions your product solves. The phrasing in your sales emails and support tickets is usually closer to how buyers talk than anything a keyword tool suggests.

For a business owner checking her own brand, that list fits on one page. For an agency operator, the same list per client becomes the backbone of monthly reporting.

Why does ChatGPT give a different answer every time you ask?

Because the model composes each answer fresh, and small changes in wording, session context, or the pages its search layer retrieves produce different results. Rand Fishkin of SparkToro found less than a 1 in 100 chance that ChatGPT or Google's AI, asked the same question 100 times, gives the same list of brands in any two of its answers. The sources move as well: Profound's study of AI search volatility, which re-ran about 80,000 prompts per platform a month apart in 2025, found that 40 to 60 percent of the domains cited had changed by the second run, even for identical questions.

The volatility also strikes at the platform level. Engines change how often they show citations at all, and when that happens, visibility moves everywhere at once for reasons that have nothing to do with any single brand. A drop in what you see is not always a drop in your content.

The practical consequence is simple: a single check tells you almost nothing on its own. What means something is the pattern across many dated checks, which is why the record you keep matters more than any one answer.

How do you check Claude, Gemini, Perplexity, and AI Overviews too?

Run the same questions through each engine, because they retrieve from different sources and disagree often. Perplexity cites sources on every answer, Claude and Gemini cite when their search tools run, and Google AI Overviews appear on a subset of searches, since Google shows one only when its systems judge it additive. A brand that ChatGPT ignores may be Perplexity's first recommendation, and the reverse happens just as often.

Keep the method identical: fresh session, same wording, note the brands, the position, and the sources. The engines lean on different crawlers and indexes, which is also why crawler access matters; the mechanics are covered in how to get ChatGPT to cite your website.

What should you record so the check is worth anything?

Record enough that a stranger could verify the check later. A screenshot alone ages badly; a structured note is better.

Field	Why it matters
Date and engine	Undated checks cannot show a trend
Exact question wording	Small rephrasings change the answer
Brands named, in order	Position is the score
What was said about you	Wrong or stale descriptions are fixable
Sources cited	These are the pages you need to influence

This is where the check becomes an asset. A dated archive of answers cannot be reconstructed later, so every week you skip is evidence you can never get back. When an agency shows a client the answer from March next to the answer from June, with the answer excerpt and cited sources for both, the before and after argues for itself.

When is a free checker enough, and when do you need daily tracking?

A free checker is genuinely enough for a one-off look. Several of the big SEO tools now bundle a basic AI visibility check into their free tiers, and for "does ChatGPT know we exist" a manual session costs nothing but ten minutes. If that is your whole question, start there.

Daily tracking earns its keep when you need to prove change over weeks. If you are publishing content to earn citations, you need before and after answers, on dates, across engines, to see whether anything moved beneath the churn. That is the layer CiteRank covers. It finds the questions your buyers ask, checks ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews, and keeps a dated record of every check: who was named, in what order, which sources were cited, and an excerpt of the answer. The trend is provable rather than remembered. The score is the share of checks that named you, which is how the daily score is calculated.

Either way, the method stays the same. Ask the real questions, in fresh chats, and write down what came back with a date on it. If the discipline behind all this is new to you, what is AEO covers the fundamentals.