To check if ChatGPT mentions your business, ask it the questions your buyers ask (not your name), each in a few different wordings and in a fresh chat, then write down whether you’re named and who is named instead.
You need no tools for this, only a notes file. It is a small AI visibility audit you can run by hand, and it shows you where you stand before you pay anyone to measure it.
What should I ask ChatGPT?
Ask the questions a buyer would type before they know your name, not “What do you know about my company?”. A question that includes your name only tests whether ChatGPT recognises the name. It says nothing about whether you come up when someone is still choosing.
Start with a short list of questions in your buyers’ words. These kinds are worth including:
- “Best” questions: “Who’s the best moving company in Portland?”
- Problem questions: “How do I move a piano up a flight of stairs?”
- Topic questions: “What should I look for in a moving company?”
These examples are made up, and the Portland question is the same illustration we use on how it works. Swap in your own city, product and customers. Open a fresh chat for each question, with no earlier conversation, so one answer doesn’t colour the next.
Why ask each question in different wordings?
We ask each question in a few wordings because, in our method, asking the same wording again mostly returns the same lookup. Our method page puts it this way: “Asking the same wording again mostly returns the same lookup, so we ask each question in 3 different wordings, once each” (Answersky, how we measure, checked Oct 3, 2026).
That is Answersky’s method, and you can borrow it for a hand check. For the Portland question, wording 1 might be “Who’s the best moving company in Portland?”, wording 2 “Which Portland movers should I hire?”, and wording 3 “I’m moving across Portland next month, who do you recommend?” It is the same need asked in 3 different ways.
Ask each wording once, in its own fresh chat. If you are named in one wording and not in the others, write that down. It is the reason we don’t treat a single answer as a verdict.
What did we find when we checked our own site?
When we checked answersky.com on Sep 30, 2026, the score was 0/100 on ChatGPT, Gemini and Perplexity, and we were named in none of the answers (Answersky /log, Entry 1, checked Oct 3, 2026).
The assistants did name businesses. They named 7 other tools that track AI visibility instead of us (Answersky /log, Entry 1, checked Oct 3, 2026). That is the useful half of the result: it shows who a buyer meets in our place.
One line on the method: we asked the models behind ChatGPT, Gemini and Perplexity, via their APIs with web search, each question in 3 different wordings, once each (Answersky, how we measure, checked Oct 3, 2026). The consumer apps can answer differently, and a hand check in the ChatGPT app is not the same measurement as our scan. Treat your own check as a first look, not a score.
The full record, including what we change from week to week, is in our public log.
Answersky baseline, Sep 30, 2026
Named in none of the answers on ChatGPT, Gemini and Perplexity
7 other tools that track AI visibility were named instead
What should I write down?
Write one row per answer, so you can compare across wordings, days and assistants. Use these columns:
- Date
- Assistant (ChatGPT, Perplexity, Gemini)
- The question
- Wording (1, 2 or 3)
- Named? (yes or no)
- Who was named instead
- Sites cited, if the assistant lists any
The “who was named instead” column matters most: it tells you which businesses a buyer was shown in your place. The “sites cited” column records the sources the assistant listed, a clue to where your market is being discussed, not proof of everything behind the answer.
Save the answer text too, or a screenshot, so you can check a row later. Answers can change, and your notes are the record of what was said on the day.
| Date | Assistant | Question | Wording | Named? | Who was named instead | Sites cited |
|---|---|---|---|---|---|---|
| Oct 5, 2026 | ChatGPT | Who's the best moving company in Portland? | 1 | No | Business A, Business B | a local directory |
| Oct 5, 2026 | ChatGPT | Which Portland movers should I hire? | 2 | No | Business B | a review site |
| Oct 5, 2026 | Perplexity | Who's the best moving company in Portland? | 1 | Yes | Business C, Business A | a review site, a local directory |
Should I check Perplexity and Gemini too?
Yes, if your buyers use them. Ask the same questions, in the same wordings, in each assistant and add the rows to the same sheet.
Perplexity’s answers list numbered sources, so you can see which sites it listed and write them in the last column. Those are the sources it listed, a clue to which pages it drew on, not proof of everything behind the answer.
We don’t claim any one of these is the biggest. Pick the ones your own customers say they use, and ask them.
Is there a free AI visibility checker?
Yes. Our free Quick Scan is one, and Bing Webmaster Tools offers a different kind of report. They measure different things.
The first is our free Quick Scan. It asks 6 of your buyers’ questions on ChatGPT and Perplexity, 12 answers, and shows whether you’re named and who AI named instead. The report arrives within 3 business days (Answersky homepage, checked Oct 3, 2026). It is a snapshot and the same idea as the hand check above, done for you. It does not replace reading the answers yourself.
The second is the AI Performance report in Bing Webmaster Tools, in public preview. Microsoft describes it as showing how publisher content appears “across Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations” (Microsoft Bing blog, Feb 10, 2026, checked Oct 3, 2026). That means Copilot and Bing AI citations, not ChatGPT, so it can’t answer this post’s question. It is another look at one family of AI answers.
How do I know if something changed?
For our own checks, we count a change only when 2 scans agree (Answersky, how we measure, checked Oct 3, 2026). That is a working rule, not proof that a change is real, and our method page says scores can vary between scans.
You can borrow the rule for a hand check: ask the same questions again on a different day, and treat a change as worth acting on only when 2 checks agree.
So keep the questions and wordings fixed, add new rows with the new date, and compare like with like. If you change a question, start a new sheet. You can read what our score does and doesn’t mean on how it works.
That is how to track AI search visibility by hand: fixed questions, a few wordings, a dated sheet and some patience. If you’d rather not run it yourself, there is a short next step.
Free Quick Scan: 6 of your buyers’ questions on ChatGPT and Perplexity, 12 answers, and who AI named instead of you. Report within 3 business days. See pricing.