Ask the questions a real customer would ask
Customers do not type your business name. They describe a problem or a need, in plain words, and ask the AI to recommend someone. So that is what you type too. Use your real category and your real city, and ask the way a person actually talks.
Start with these three. Swap in your own trade and town.
- best [category] in [city] (for example: best emergency plumber in Tucson)
- [category] near me open now (set your location, or name the neighborhood)
- who should I call for [specific problem] in [city] (for example: who should I call for a leaking water heater in Tucson)
Add the follow-up questions customers really use
The first three give you the broad picture. The next set is where the money is, because they match how people decide. People do not just want a name, they want a reason: cheapest, highest rated, open late, speaks their language, handles their exact job.
Run a few of these too, again with your category and city filled in:
- affordable [category] in [city] with good reviews
- [category] in [city] open on weekends
- most recommended [category] near [neighborhood]
- [category] in [city] that does [your specialty] (for example: dentist in Tucson that does same-day crowns)
Ask each question more than once, and count
Here is the part owners get wrong: they ask once, see they are missing, and panic, or they ask once, see their name, and relax. Both are mistakes. AI answers are not fixed. The same question, asked three times in a row, can give you three different lists. The model is sampling, not reading from a fixed ranking.
So do not treat one answer as the truth. Ask each question at least three times, in fresh chats, and count how often you appear. What matters is the rate, not a single hit. Showing up in one run out of three is very different from three out of three, and only the rate tells you where you really stand.
Open a new chat each time, not a follow-up in the same one, so the AI is not just repeating what it already said. A quick way to track it: write the question down, then put a tally mark each time your name appears across the runs.
Check all four places, not just ChatGPT
ChatGPT is the one people name, but your customers use more than one tool, and the four big ones do not agree with each other. A business that ChatGPT loves can be invisible on Perplexity. So check all four, running the same questions in each.
Where to run them:
- ChatGPT (chatgpt.com), with web search on so it pulls live local results
- Gemini (gemini.google.com), which leans hard on Google's own local data
- Perplexity (perplexity.ai), which shows its sources, so you can see where it got its facts
- Google AI Overviews: do a normal Google search for your questions and read the AI summary box at the top of the results
Write down three things for every answer
Reading the answers is not enough. You need a record you can act on, because patterns only show up once you have it written down. Keep it simple. For each question, in each tool, note three things.
Do not skip the second one. Wrong facts are often more damaging than being left off entirely, because the AI states them with full confidence. If it says you close at 5 when you are open until 9, or tags you as permanently closed after an old move, you are losing customers who never even call to check.
- Where you appear: are you named, and are you near the top of the list or buried at the bottom
- What they get wrong: copy the exact wrong fact word for word, wrong hours, wrong price, a 'closed' flag, a service you do not offer
- Who they recommend instead: write down the competitor names that come up, and any reason the AI gives for picking them
Read the competitor reasons, because that is your to-do list
When you note who the AI recommends instead of you, also catch why. The reasons are usually right there: 'highly rated with over 200 reviews', 'known for fast emergency service', 'open 24 hours'. Those phrases are the AI telling you, in plain English, what it rewards.
If the same reason shows up for several competitors and never for you, that is your single biggest gap. Maybe it is review volume. Maybe it is that nobody has written that you handle emergencies. You now know what to work on, and you did not have to guess. For the full playbook on closing those gaps, see the guide How to get your local business recommended by ChatGPT, and if you want to understand the broader practice, read What is GEO (Generative Engine Optimization)?
Turn it into a score, not a feeling
Now total it up. Add the times you appeared and divide by the times you asked. If you ran 4 questions, 3 times each, across 4 tools, that is 48 chances. Showing up 12 times is a 25 percent rate. That number is your real visibility, and it beats any single answer or gut feeling. Save it. In a month, after you have made changes, run the exact same questions again and compare the rate. That is how you prove a fix actually worked instead of hoping it did.
Doing this by hand is real and it is free, and every owner should run it at least once to see their own situation with their own eyes. It is also tedious. Three runs across four tools for a handful of questions is dozens of chats, and the numbers shift while you type. This is exactly what LocalFox automates: it asks the customer-style questions across ChatGPT, Gemini, Perplexity and Google AI Overviews, three runs each, quotes the wrong facts word for word, lists which competitors get recommended and why, and gives you a visibility score plus a copy-paste fix kit. The free check shows your score and your single biggest problem with no account. The manual version above is the same idea, just done by hand. If you are not appearing at all and want to understand why before you spend a cent, read Why isn't my business showing up in AI search?