SeeWAIS
2026-09-25

ChatGPT SEO audit for local business: the five-step self-check

Five steps to check what AI tells buyers about your local business — and the honest line on what a self-check can and cannot fix.

Before a buyer calls a plumber, a med spa or a dental practice, a growing share ask an assistant first. What that assistant says — whether it names you, whose price it quotes, whose story it tells — is already part of your local marketing, whether you have ever looked at it or not. The good news: you can look at it yourself, tonight, for free. Five steps, an evening, a notes file. Then the honest line on what a self-check finds versus what it takes to fix.

Step 1: ask who it recommends, and see whether you exist

Open a fresh chat and ask the question a buyer would ask — without your name in it:

Who are the best AC repair companies in San Antonio?

That is one of the real questions from our measured runs (an HVAC contractor in San Antonio). A med spa version: "Where can I find a med spa in Nashville that offers both Botox and dermal fillers?" Then record three things, not one:

  • Were you named — and was it a recommendation ("start with X"), a mention in passing, or a citation while someone else got picked?
  • Who got named instead, and in what order?
  • What reason did the assistant give — price, reviews, a specific service page?

One question per new chat. Anything asked earlier in the same conversation contaminates the next answer, and putting your own name in the question defeats the test — the point is whether the assistant produces your name unprompted. The full DIY procedure, including what to record, is in our guide to checking properly.

Step 2: ask what things cost, then check the number against reality

How much does Botox cost in Nashville?

That is a real question from our published sample run. Assistants answer cost questions even when your own prices are unreadable to them: the number gets assembled from whoever published a machine-readable figure — a directory, an aggregator, a competitor's price menu. In that sample run (a real client, pseudonymised), AI produced 53 price quotes about the business: 36 matched the business's own published prices exactly, and 17 had no price on the site to check them against. Whatever figure comes back for your services, check it against what you actually charge. An inflated number kills the inquiry before it happens; a suspiciously low one starts the relationship as a bait-and-switch. We wrote up the whole mechanism in why ChatGPT quotes the wrong price for your business.

Step 3: ask what AI knows about you, and check the identity

Now, and only now, use your name: "What do you know about [your business]?" You are looking for two failure modes:

  • Wrong facts — services you do not offer, practitioners who do not work there, hours that drifted.
  • An identity that cannot be confirmed. In our published sample — a real client run, pseudonymised — the business ranked #1 of 168 candidates and still carried 14 wrong facts, including several answers that said plainly the assistant could not confirm the business exists or identify its location, services or prices. Rank first, unable to be confirmed.

That pair of findings is worth sitting with: a ranking and a correct identity are two different measurements, and a good rank can hide an identity that never got confirmed.

Step 4: check whether AI is allowed through the door

If the crawlers are blocked, nothing else matters — a system that cannot read you cannot quote you. Open /robots.txt on your domain and look for the AI crawler directives: GPTBot (OpenAI), ClaudeBot (Anthropic), Google-Extended, PerplexityBot. In the sample run we checked 13 such crawler directives and not one was blocked — that client's problem was never the gate. Then look one layer down: do the pages render without JavaScript, does the structured data actually parse, is there a stray noindex anywhere. This is the cheapest layer to fix and the first thing any honest audit checks. Our free check runs this part automatically.

Step 5: write it down — today is your baseline

Whatever you found: paste the full answers into a file with the date, and ask the identical questions again on a different day. Assistants that search the web do not give the same answer twice; one run is an anecdote, and the drift is itself a finding. Without a written baseline you will never be able to tell, months from now, whether any of this moved.

What the self-check finds — and what it cannot fix

An evening of this tells you whether you exist to AI, roughly whether the story is right, and whether the gate is open. That is genuinely worth having, and it is where most businesses should start.

What it cannot do is tell you why — which sites the answers were built from, and whether any of them mention you — or repair the story systematically. Two of our measured examples show the gap. The pseudonymised med spa in the published sample ranked #1 of 168 and still needed fourteen wrong claims repaired one by one. A family dental practice we measured in Iowa scored 49.8 out of 100 (Emerging), ranked #1 of 85, with only two wrong facts out of 277 checked — and its one serious problem was invisible to any ranking: a membership program the practice runs, which one assistant told a buyer did not exist. Both businesses were #1. Only the claim-level work tells them apart. The full breakdown of that practice is on our dental industry page.

A systematic repair looks like this: 30 real buyer questions across five AI surfaces (176 graded answers, 302 API calls) — the method is published and frozen; every checkable claim fact-checked against your live pages; a fix list sorted by who on your team can execute it; and three copy-paste blocks — an llms.txt, structured data, a disambiguation paragraph — generated from your own run. Then ask the same questions again a few weeks later, to see whether the wrong claims actually cleared. What it costs: $4.99 once, no subscription.

Start with the free check

Run the five steps above this week — they cost an evening. And run the free check first: it tells you whether AI systems can technically read your site, and shows you the exact questions a full measurement would use — all 30, in full — before you pay anything or even make an account.

What does AI say about your business?

The free check reads your site and shows you the exact questions we would ask — before you enter any card details. The full report is $4.99 once.

Keep reading: