How to check if ChatGPT recommends your brand

This is the manual version of tracking your brand in AI search. It costs nothing but time, and it is worth doing once by hand even if you later automate it, because it shows you what the data actually looks like before you pay anyone for it.

Step 1 — write the questions

You need the questions a buyer asks before they know your brand exists. Not "is [your brand] any good" — nobody who hasn't heard of you asks that. The useful ones look like:

• "what kind of tool tracks whether AI mentions a company"
• "best [category] tool for a small agency"
• "alternatives to [a competitor] for [use case]"
• "how do I [the problem your product solves]"

Ten to twenty of these. Keep your own brand name out of them — you are testing whether ChatGPT brings you up unprompted.

Spread them across the buyer's journey. A few "does this kind of thing exist" questions, a few "best option for my situation" questions, a few "how do I actually do this" questions. Brands tend to be visible at one stage and absent at another, and a panel weighted to one stage hides that.

Step 2 — decide what you are recording

For every answer, four columns:

Question	Named	Recommended	Others named
best tool to see if AI mentions a brand	yes	listed, 4th of 6	Profound, Otterly, Peec AI
how do I check my brand in ChatGPT	no	—	Semrush, a Reddit thread

• Named — was your brand mentioned at all, yes or no.
• Recommended — singled out as a choice, listed in passing, or actively steered away from. "Named" and "recommended" are different results and you want both.
• Others named — every competitor that appeared, in order.

The last column is what turns this from a vanity check into something useful. After twenty questions you can see which rivals own the category in ChatGPT's view, and on which kinds of question they own it.

Step 3 — ask each question several times

ChatGPT's answer to the same question is not fixed. It varies from one session to the next, and more so when it runs a live web search. Ask each question at least three times, on different days or in different sessions, and treat the majority outcome as the reading. If two of three answers name you and one does not, you are "named" — but note the variance, because a brand mentioned inconsistently is in a weaker position than the count alone suggests.

Use a temporary chat or log out first. A logged-in session carries your history, and if you have been researching your own category, ChatGPT may lean toward brands you have already looked at, which flatters the result.

Step 4 — run it for your competitors

Do the whole thing again for two or three rivals, same questions. This is not optional. "Named in 6 of 20" is meaningless until you know the leader was named in 18 and the nearest competitor in 11. Your position is a share, not a count, and the share is the only number worth reporting to anyone.

Step 5 — read the result

Three figures come out of the log:

• Mention rate — questions that named you, over questions asked.
• Recommendation rate — questions that recommended you, over questions asked. This is always lower, and the gap between the two is a real signal: named often but recommended rarely means the engine knows you exist but has no reason to prefer you.
• Share of voice — your mentions as a fraction of all brand mentions across the panel.

Write them down with the date. That row is your baseline.

Step 6 — repeat on a schedule

One pass is a baseline. The value is in the second and third passes a few weeks apart, once you have published something or earned a mention on a page ChatGPT reads. Keep the questions identical between passes or the comparison breaks. This is also the step that decides most people against doing it by hand indefinitely — 15 questions, 3 engines, 3 repeats, every week, is around 135 answers to read and log without missing a week.

What this does not tell you

Whether your brand was named is the what. It does not tell you why — that lives in the sources behind each answer, which ChatGPT only sometimes shows. If you want the diagnosis and not just the score, that is the next thing to look at. And before you trust the score itself, read why AI visibility scores move between runs.