Why one chat tells you nothing
Most people check their ChatGPT visibility by typing a question they think is flattering, in an account that already knows them, once. Whatever comes back, they believe it. The problem is that ChatGPT’s answers vary: between sessions, between accounts with different history, between locations, and between the retrieval and no-retrieval paths it chooses per question. A single answer is a draw from a distribution. To know where you stand you need several draws, taken the same way, and the way has to be written down so that next month’s check is comparable.
A method you can repeat
1. Write the questions buyers ask
Not the questions you wish they asked. Use their words, including the sloppy ones. A useful set mixes five types:
- Awareness: “what should I look for in a payroll provider for a small Estonian company?”
- Comparison: “X vs Y for …”, “alternatives to X”
- Commercial: “best …”, “how much does … cost”, “who can … by next week”
- Local: the same questions with a city or country attached
- Brand: one or two with your name in them, to see what ChatGPT believes about you
Ten to fifteen questions is enough to see patterns. The commercial and comparison ones are the ones that decide purchases, so weight your reading toward those.
2. Ask them the way a buyer would
Logged out, or in a fresh account with no memory and no custom instructions. Add one neutral line of context if your market is local (“I am asking from Estonia”), since retrieval varies by location. Do not add anything that steers toward your brand. Ask each question in a new chat, and ask each question at least twice on different days.
3. Keep everything
Paste the full answer into a document with the date, the question, and the sources shown (if any). Mark three things per answer: is your brand named; is your website among the sources; and what, if anything, does the answer claim about you. Then mark every competitor named. After ten questions run twice, you have twenty answers and a first picture.
4. Check the gate
Open your robots.txt. OpenAI documents three agents: one used for training the models, one that indexes for ChatGPT search, and one that fetches pages when a user asks. If the search agent is blocked, your pages cannot be sources in ChatGPT answers, however good they are. This is the single most common reason a company is named but never cited.
How to read what you find
Named and cited. ChatGPT knows you and trusts your site enough to use it. Check the accuracy of what it says; small errors spread.
Named, not cited. ChatGPT is learning about you from someone else: directories, review sites, press, a competitor’s comparison page. Find those sources in the answers. If they are wrong about you, fix them at the source. Then make your own pages the easier thing to cite: readable without JavaScript, answer-first, factual.
Absent. Look at who is named instead and where ChatGPT cites them from. Those pages define the category in the engine’s view. Being present on the same kind of pages, and having your own page that answers the same question plainly, is the route in.
Wrong. A wrong price, a wrong city, a service you do not offer. Treat these as the first fixes: correct the fact on your site, on the third-party pages ChatGPT cited, and anywhere else it is stated. The engine repeats what the most-cited sources agree on.
What this method cannot tell you
It cannot tell you how many buyers asked. It cannot give you a market-wide share of voice; the counts are counts within your sample. It cannot predict next month, because the engine changes. And it cannot be compared with a check done a different way, which is why the method matters more than any individual result.
Doing it at scale
The manual version takes an afternoon for one engine. The same check on ChatGPT, Perplexity, Google AI Overviews and Claude, with repeat runs from your market, the robots.txt and page-citability checks run by code, the claims compared against your facts, and a person reading the result, is what our audit does. If you only want to know whether there is a problem, the free check runs three buyer questions on ChatGPT and Perplexity and emails you the answers with your brand marked. The methodology describes the procedure in detail; the guide on getting mentioned by ChatGPT covers what to do with the findings.