The method, step by step
You can audit yourself in about twenty minutes. The point is to see what a real prospective customer sees — so control for the things that bias the answer.
- Use a clean session. Open ChatGPT in a temporary or logged-out chat (so your history and memory don’t flatter you), and use a private/incognito window for Gemini and Perplexity. A personal, logged-in account contaminates the result.
- Ask the customer’s question, not your name. Type what a prospect types before they’ve chosen anyone: “best [your service] in [your city],” “affordable [service] near me,” “who should I hire for [need].” Asking your business name only tests your reputation, not whether you get recommended.
- Check all the engines your customers use. At minimum ChatGPT, Gemini, and Perplexity; ideally also Google’s AI Overviews (the AI box on a normal Google search) and Google’s AI Mode. They run on different sources and will disagree.
- Run each question two or three times. Note who gets named, in what order, and whether your business appears at all.
- Write down the sources. Where an engine shows citations (Perplexity always does), note which directories, reviews, and pages it pulled from. That’s your fix list in raw form.
What to look for
- Are you named at all? If competitors come back and you don’t, that’s the gap.
- Reputation vs recommendation. Ask “is [your business] reputable” and you’ll likely get a glowing answer. Ask “who’s the best [service] in [city]” and you may vanish. The contrast is the whole problem — you can be praised on demand and invisible when it counts.
- Which engine, which question. You’ll often win one engine and lose the rest. That’s normal — each runs on a different source stack.
The catch that trips everyone up: ask the same question twice and you’ll get two different answers. AI engines are non-deterministic, so the businesses they name change run to run. Do not make decisions off a single check — it’s a coin flip. Run each question at least three times and look at the pattern, not one result.
When to stop DIY-ing
A manual pass tells you roughly where you stand — enough to know whether you have a problem. It won’t give you a defensible rate, per-engine scoring, or a prioritized fix list, and it’s easy to fool yourself with a lucky run. When you want the honest number — 25 questions, three runs each, six engines, every result logged — that’s how we measure it, and the first report is free. Either way, once you know where you stand, the next step is making your business show up in AI.
The rigorous version
Then fix it
Common questions
How do I see what ChatGPT says about my business?
Open a fresh, logged-out or temporary ChatGPT session (so your history doesn't bias the answer), and ask the questions your customers ask before hiring — for example 'best [your service] in [your city]' — rather than your business name. Note whether you're named and who's recommended instead. Repeat the question two or three times, because the answer changes run to run.
Why do I get a different answer every time I ask?
AI engines are non-deterministic — they generate a fresh search and compose a new answer almost every time, so the businesses they name shift between runs. That's why a single check is unreliable, and why any honest measurement reports a rate across several runs rather than one screenshot.
Should I ask the AI about my business by name?
Ask both ways. Asking 'is [business] reputable' tells you your reputation answer; asking 'who's the best [service] in [city]' tells you the recommendation answer. Many businesses are praised when asked about by name and invisible when customers ask who to hire — and the second one is what loses you the client.