Quick answer: you can find out exactly what AI assistants tell customers about your business in about 15 minutes, using four types of prompts asked across ChatGPT, Perplexity, and Gemini — and repeating each one a few times, because AI answers change between runs. What you learn falls into four patterns (described wrong, competitor recommended instead, not known at all, or accurately recommended), and each pattern points to a different fix. This guide gives you the exact prompts, what to look for in the answers, and what each result means — no tools required.
Most owners have Googled themselves. Almost none have ChatGPT'd themselves — even though a growing share of their customers ask AI assistants for recommendations before they ever reach a search results page. This is the new "Google yourself," and it takes a quarter of an hour.
Why check what AI says about you?
Because AI assistants are now answering questions about your business whether you participate or not. When someone asks "is [your business] any good?" or "best [your category] near me," the assistant composes an answer from whatever it knows and retrieves — your site, directories, reviews, forum threads, articles. If that picture is wrong, stale, or empty, the AI presents its version confidently to a customer you never see. The only way to know which version of you it's serving is to ask it yourself.
Before you start: two rules that make the check valid
Rule 1 — Ask each prompt 3 to 5 times, in fresh chats. AI answers are non-deterministic: the same question produces different answers on different runs. Ask once and you have an anecdote; ask five times and you have a pattern. (You'll see this yourself by run three — the brand lists shuffle. That's normal, and it's why any tool selling you a fixed "AI rank" should be treated with suspicion.) Use a new conversation for each run so earlier answers don't contaminate later ones.
Rule 2 — Ask like a customer, not like the owner. Log out or use a browser profile that isn't soaked in your own brand searches where you can. Phrase prompts the way a stranger would — "best plumber in Rochester that does emergency calls," not your internal category jargon.
The four prompts (run each on ChatGPT, Perplexity, and Gemini)
Prompt 1: The identity check — "What is [your business name] in [city]?"
You're testing whether the AI knows you exist and describes you correctly. Read the answer for: right category? right location? current offerings, or the menu you retired two years ago? anything invented (a hallucinated founding date, a service you don't offer)?
Prompt 2: The recommendation check — "What's the best [your category] in [your city/area]?"
The money question — the one your customers actually ask. Note every business named, in a simple tally across your runs. You're measuring three tiers: are you mentioned at all, cited (linked), or actively recommended? Also note who shows up instead of or alongside you — that's your real competitive set in this channel, and it often differs from who you think your competitors are.
Prompt 3: The comparison check — "Alternatives to [the competitor who came up most]" and "[your business] vs [that competitor]"
This reveals how the AI positions you when a customer is actively choosing. Does it know your strengths? Does it repeat one old review's complaint as if it were your defining trait? Comparison prompts are where inaccurate or thin information hurts most, because the customer is at the decision point.
Prompt 4: The trust check — "Is [your business name] trustworthy? What do reviews say?"
AI assistants synthesize review sentiment into a verdict. You're checking which sources it leans on (Google reviews? Yelp? a forum thread from 2021?) and whether the summary matches your actual current reputation. On Perplexity especially, note the cited sources — those exact pages are where your reputation lives, in the AI's eyes.
How to read your results: the four patterns
Pattern A — Described wrong or out of date. The AI knows you but has stale or invented details. Fix: entity cleanup. Make your homepage state plainly what you are, update your About page, and align every directory and profile to one consistent name, category, and description. Inconsistency across the web is what blurs the AI's picture of you.
Pattern B — Competitor recommended instead. You exist but lose the recommendation. Fix: find out where the AI's answers come from — note the sources it cites for your category prompts, open them, and check whether you appear in those pages. Being absent from the listicles, directories, and review platforms the AI already cites is the most common and most fixable cause. Full article: Why does ChatGPT recommend my competitor instead of me?
Pattern C — Not known at all. The AI has effectively never heard of you. Check the plumbing first — can AI crawlers access your site? — then build the basics: a clearly-described website, presence on major directories and review platforms, and content that answers the questions your customers ask. Full article: Is your website visible to ChatGPT?
Pattern D — Accurately recommended. Congratulations — now protect it. Note the sources being cited on your behalf and keep them healthy (fresh reviews, current listings), because the AI's picture of you is only as durable as its sources.
Make it quarterly (and write it down)
A single check is a snapshot. The value compounds when you repeat it: keep a simple sheet — date, prompt, engine, mentioned/cited/recommended, how described — and run the same prompts every quarter. What you're building is a trend line, and the trend is the only honest measurement in a channel where individual answers vary. Fifteen minutes, four times a year, and you'll know something most of your competitors don't: what the machines that increasingly introduce businesses to customers actually say about yours.
If you'd rather not do it by hand: this exact loop — same prompts, multiple runs, across engines, tracked over time — is what SeeGeo automates, alongside the site-side checks (crawler access, content structure) that the manual method can't see. But run the manual check at least once regardless. Reading the AI's own words about your business, live, is worth more than any dashboard for making the channel feel real.
Frequently asked questions
What should I ask ChatGPT to see what it knows about my business? Four prompts, asked like a customer: "What is [business name] in [city]?" (identity), "Best [category] in [city]?" (recommendation), "[Business] vs [competitor]" (comparison), and "Is [business] trustworthy — what do reviews say?" (trust). Run each 3–5 times in fresh chats on ChatGPT, Perplexity, and Gemini, and tally the pattern rather than trusting any single answer.
Why does ChatGPT give different answers each time I ask? AI answers are non-deterministic by design — the same question legitimately produces different responses across runs. That's why a valid self-check repeats each prompt several times and reads the pattern, and why single-run "AI rankings" are noise.
ChatGPT described my business incorrectly. Can I fix it? Yes, indirectly. AI systems learn from your website and everything published about you, so the fix is making the correct information unmissable: a plain-language description at the top of your homepage, an accurate About page, and consistent name/category/location across every directory, profile, and review platform. Corrections propagate over weeks to months depending on the engine.
Can I pay to be recommended by ChatGPT? No — there's no buyable placement in the organic answers of the major assistants as of 2026. Recommendations come from being accessible to their crawlers, clearly described, and present in the sources they cite. That's earned, not bought — which is bad news for shortcuts and good news for small businesses that do the work.
How often should I run this check? Quarterly as a baseline, plus after anything that changes your public footprint: a rebrand, a move, a site migration, a review surge, or a burst of press. AI models and their sources update continuously, so an annual check misses the drift.