There is a test. It takes about an hour, costs nothing, and gives you a number you can watch move. The hard part is not running it — it is refusing the shortcut that makes it meaningless.
Why searching your own brand name proves nothing
Type your company name into ChatGPT and it will describe you, probably well. That feels like evidence. It is not. You supplied the name, so the model is working around a term you handed it. A potential customer never does that — they have a problem, not your name.
The only question worth measuring is whether you come up when nobody mentions you.
The test, in five steps
- Write twenty questions a customer would actually ask. Not "best CRM software" — the awkward, specific things people really type: "how do I stop losing track of quotes when three people answer the same client".
- Open a session with no memory of you. A logged-out window, or a temporary chat. If the assistant remembers previous conversations about your company, the result is worthless.
- Ask each question in three assistants. ChatGPT, Gemini and Perplexity retrieve differently and disagree more than people expect. One of them naming you and two ignoring you is itself a finding.
- Record three things per answer: were you named, who was named instead, and which sources were cited.
- Add it up. Twenty questions across three assistants is sixty answers. Your citation rate is how many of the sixty include you.
What to write down
| Question | Assistant | Named you? | Named instead | Sources cited |
|---|---|---|---|---|
| Question 1 | ChatGPT | no | Competitor A, Competitor B | a review site, a forum thread |
| Question 1 | Gemini | no | Competitor A | a trade publication |
| Question 1 | Perplexity | yes | — | your documentation page |
The third column is the score. The fourth and fifth are where the actual work comes from: the sites being cited instead of you are a list of exactly where you need to exist.
Reading the number
- 0 of 60. You are invisible, not unpopular. Start with references, not with copywriting.
- 1 to 6. You exist but you are not a default. Usually means a single strong source is carrying you.
- 7 to 20. You are in the consideration set. Now wording matters: check whether you are described correctly.
- Above 20. You are a default answer in your category. The job becomes defending it.
How often to run it
Monthly, with the same twenty questions. Changing the questions each time feels more thorough and destroys the comparison. Keep the list frozen for at least a year; add a second list if you want to track something new.
Do I need a tool for this?
Tools exist and they automate exactly this. They are worth paying for once you are tracking it continuously. Run the first round by hand anyway: reading sixty answers teaches you how competitors are described, which no dashboard will hand you.
