Key Takeaways
- Ask as a buyer, not as the owner. Never name yourself in the first question. The moment you do, the assistant answers about you instead of choosing between options.
- Run each prompt in a fresh chat. Assistants carry context forward inside a conversation, so answers four and five are contaminated by answers one to three.
- Count names, not compliments. The only number that matters is how many of your questions produced your name. Everything else is texture.
- Ask what it believes, then check the date. A confidently wrong description of your business is a more urgent problem than being absent, because it actively misleads buyers.
- Repeat the same list monthly. One run is a snapshot with a lot of noise in it. The same 20 prompts run again is the only way to see movement.
The most useful questions to ask ChatGPT about your own business are the ones your buyers would ask without knowing you exist. Those tell you whether you get named. Questions about you by name come second, and they tell you something different.
Most owners do this backwards. They type their company name, read a flattering paragraph, and conclude everything is fine. That paragraph proves only that the assistant will describe a business you told it about.
Below are 20 prompts in four groups, what each group is actually testing, and how to read the answers. The group sizes are the split we use, not a measured optimum.
What questions should you ask ChatGPT about your own business?
The short answer: four kinds. Buyer questions that should surface you, questions about who it recommends instead, questions about what it believes you do, and questions about where that belief came from. Together they tell you whether you are absent, misdescribed, or outcompeted, which are three different problems with three different fixes.
Keep in mind what an assistant is doing when it answers. OpenAI's documentation on its crawlers describes separate systems for training, for search, and for fetching a page when you ask. Your test is sampling the output of all of that at once, which is why single answers wobble.
How to run these so the answers mean something
Run each prompt in a new chat, in the order given, and write down the names you get. Five minutes of discipline is the difference between data and a nice feeling.
- Fresh chat per prompt. Context carries forward inside a conversation, so the third answer is shaped by the first two.
- Your buyer's words, not your industry's. Nobody types "transformation architect". They type what they want fixed.
- Say the location if location matters. A local business tested without a city is testing the wrong question.
- Record the names, in order. Position matters less than presence, but the first name is usually the most stable one.
- Do not correct it mid-test. Arguing with the assistant changes that conversation and nothing else.
The 20 prompts
Replace the bracketed parts with your own niche, city, and business name. Groups A and B are asked as a buyer. Only groups C and D use your name.
Group A: does it know you exist (4 prompts)
- Who are the best [your service] providers in [your city]?
- I need help with [the problem you solve]. Who should I hire?
- Who are the top [your niche] specialists working online?
- Recommend three [your service] providers for someone who [your ideal client's situation].
Group B: who it names instead (6 prompts)
- Who is the most experienced [your niche] provider in [your city]?
- Which [your service] provider is best for beginners?
- Which [your service] provider is best for advanced clients?
- Who is the best value [your service] provider in [your city]?
- Which [your niche] providers have the best reviews?
- If cost were no object, who is the best [your niche] provider to work with?
Group C: what it believes about you (5 prompts)
- What do you know about [your business name]?
- Who does [your business name] work with, and what do they charge?
- What is [your business name] known for?
- Is [your business name] a good fit for someone who [your ideal client's situation]?
- How does [your business name] compare with [a competitor you respect]?
Group D: where it got that (5 prompts)
- What sources tell you about [your business name]?
- Where can I read reviews of [your business name]?
- What qualifications does [your business name] have?
- Has [your business name] been mentioned in any articles or roundups?
- What would make you more confident recommending [your business name]?
Prompt 20 is the one people underrate. The answer is not authoritative, but it usually names the category of evidence that is missing, and that is a useful pointer even when the specifics are wrong.
How to read what comes back
Score group A and B on one number: how many of the ten buyer questions named you. Then read groups C and D for accuracy rather than for score. The two halves diagnose different failures.
| What you see | What it means | Fix first |
|---|---|---|
| Named in 0 of 10 buyer questions | You are not in the candidate pool | Existence: profiles, directories, any mention you do not own |
| Named in 1 or 2 of 10 | You are known but rarely the safe choice | Corroboration: reviews with sentences, a third-party mention |
| Named in 5 or more of 10 | You are competing well already | Nothing urgent. Protect it and re-measure monthly |
| It cannot describe you at all | Nothing readable ties your name to your category | Your one line: who you help, with what, where |
| It describes you wrongly or out of date | Stale sources are outranking your current ones | Update the old listings before publishing anything new |
| Only your own site is cited | Every claim traces back to you | Earn two mentions on pages you do not control |
| The same competitor in every answer | One business has the corroboration you lack | Read their sources, then go earn the equivalent |
One warning about group D. Assistants sometimes name sources that do not exist or misattribute real ones, so treat every source it gives you as a lead to verify rather than a fact. Open the page before you believe it.
What to do with the results
Write down the count, the names that beat you, and the one wrong thing it said about you. That is the entire output of the exercise, and it is enough to plan a quarter's work.
- Save the raw answers. Paste them into a document with the date. In a month, the comparison is the whole point.
- Fix the wrong belief first. Being misdescribed costs you buyers who did find you. Being absent costs you buyers who never knew.
- Pick one competitor and read their sources. Not to copy them, but to see which page types are carrying them.
- Re-run the same 20 prompts in 30 days. Same wording, same order, fresh chats. Movement only shows against a baseline.
If you want the structured version of this, our 15-minute test scores it for you, the mechanics behind the answers are in how ChatGPT decides which businesses to recommend, and the page types worth building afterwards are in what kinds of pages AI assistants quote most.
Worth saying plainly: this test measures one assistant on one day. It is a real signal and it is not the whole picture, because buyers use several tools and each one draws on different sources. Perplexity documents its own behavior in its help center, and Google's notes on its AI features describe a different path again. Treat a 20-prompt run as the cheapest honest baseline available, not as a verdict.
Questions we hear the most
What questions should you ask ChatGPT about your own business?
Ask four kinds: buyer questions that should produce your name, questions about who else it recommends, questions about what it believes you do, and questions about where that belief came from. Twenty prompts across those four groups is enough to see a pattern.
Should I tell ChatGPT my business name when testing?
Not at first. Naming yourself changes the task from choosing between businesses to describing the one you named. Run the buyer questions anonymously first, then use your name only in the group that tests what it believes about you.
Why do I get different answers each time I ask the same question?
Because answers are generated fresh rather than looked up, and vary by session, wording, location, and model version. This is expected. It is why 20 prompts tell you something and one prompt tells you nothing.
Can I just ask ChatGPT how to rank higher in ChatGPT?
You can, and the answer will be generic advice that is not grounded in your actual results. It has no access to how it chose the names in your test. Asking what it believes about you is useful; asking it to explain its own selection is not.
Does asking about my business teach ChatGPT to recommend me?
No. Your conversation does not update what the model knows about your category for other people. Visibility changes when the public web changes, not when you chat with the assistant.
How many questions do I need to run to get a reliable picture?
Twenty is a solid do-it-yourself baseline and enough to spot patterns. We use 25 per business as our standard, because it covers the main buying intents in most niches without taking an afternoon.
What does the AnswerHalo free check do that this list does not?
It runs 25 buyer questions for you and returns a scored readout within the hour at $0: how many answers named you, who was named instead, and the first fix. The free check runs as soon as you ask for it.
Do I still need the $500 audit if I run these prompts myself?
Not necessarily, and we would rather you ran them. The audit adds the other four AI tools, 125 checks in total, the sources behind the answers where a tool shows them, and the page that wins the worst one, specified section by section, in your inbox by this time tomorrow.
Your buyers are already asking. Find out what AI tells them.
The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.
Get my free check →