Iron Goo - AI upgrade services for SMBs

---
title: "How to Check Whether AI Already Knows Your Business"
seoTitle: "How to Check Whether AI Knows Your Business"
description: "A twenty-minute audit any owner can run to see what ChatGPT, Gemini, and AI search already say about the business, and exactly where the gaps and errors are."
datePublished: "2026-06-24T15:17:00Z"
dateModified: "2026-06-24T15:17:00Z"
category: ai
imageAlt: "Iron Goo blog featured image on an owner running a short audit to see what AI platforms say about their business."
tags: [ai-search, ai-audit, smb-ai, entity-seo, aeo]
faq: true
---

Open a chat window right now and type your own business name, then "what do they do and where are they." Read what comes back slowly. Some of it will be right. Some line will be wrong in a way that makes your stomach drop: an address you left two years ago, a service you stopped offering, a sentence that has quietly merged you with another company that shares half your name. The real question is not whether AI is impressive. It is whether **does ai know you** the way your customers need it to, and whether you have ever actually sat down and checked. Most owners never look. They assume AI either knows them or does not, like a light switch. The truth is more specific than that, and you can find it in about twenty minutes.

I have sat next to a lot of owners while they ran this for the first time. The face change is always the same. There is a half-second of pride when the assistant clearly knows the company exists, then a slower, more useful reaction when they notice the thing it got wrong and realize a stranger asking about their trade would have read that same wrong thing. None of it is mysterious. An AI platform describes your business from whatever it could find about you. If what it found is stale or scattered, the description is stale or scattered. You can see the gaps yourself. You just have to ask the right small set of questions and look at the answers honestly.

## How can I check what AI says about my business?

Ask several AI platforms (ChatGPT, Claude, Gemini, and one AI search) the same short questions about your company, then sort every answer into right, wrong, or missing. Running identical prompts across more than one platform, not just one, is the whole audit. About twenty minutes.

That is the forty-word version. The rest of this is how to run it so the result is actually useful: the exact questions, why you must use more than one platform, and how to read what comes back without either panicking or trusting it.

## Why one platform is not enough

The single biggest mistake here is checking one place and stopping. Someone asks ChatGPT about their business, sees a decent answer, and concludes AI "knows them." Then a customer asks Gemini the same thing and gets a stale address, and the owner never finds out.

Different AI platforms were trained and connected to live sources differently. They do not read the same things, and they do not agree with each other about you. That disagreement is not noise to ignore. It is the most valuable signal in the whole exercise. When one platform has your current services and another lists a service you dropped, you have just learned that somewhere out there a stale source is still telling the old story, and at least one platform is reading it. A single-platform check hides exactly the thing you came to find.

So the rule is simple and not negotiable: run the same questions across at least three AI platforms. Treat ChatGPT, Claude, and Gemini as your core set, and add an AI-powered search tool if you use one. You are not looking for one verdict. You are looking at a spread of answers and reading the differences between them.

## The twenty-minute audit, step by step

Here is the procedure. It is short on purpose. You are not trying to be exhaustive today; you are trying to see clearly.

- **Write your five questions once.** Use the same wording on every platform so the answers are comparable. The five that matter: *"What is [exact business name] and what do they do?"*, *"Where is [business name] located and what areas do they serve?"*, *"What services does [business name] offer?"*, *"Is [business name] the same as [a similarly named company, if one exists]?"*, and one buyer question with no name attached, like *"who would you recommend for [your trade] near [your town]?"*
- **Open three platforms in three tabs.** ChatGPT, Claude, and Gemini, plus an AI search if you have one. Sign in normally. You want the everyday experience a customer would get, not a special mode.
- **Ask all five on each, and capture the answers.** Paste each response into one document, labeled by platform. Screenshots work too. The point is to have all the answers side by side, because the comparison is where the insight lives, not in any single reply.
- **Run the no-name buyer question last on each.** This is the one that tells you whether you show up at all when the customer does not already know your name. Note whether you are mentioned, mentioned wrong, or absent while a competitor gets named.

That is it for the doing. Fifteen to twenty minutes if you do not get distracted. One discipline matters while you do it: keep the wording identical on every platform. Resist the urge to reword a question to help a platform that stumbled, because identical prompts across ChatGPT, Claude, and Gemini are what let you compare the answers and spot which platform is reading a stale source. Now comes the part most people skip, which is reading the result properly.

## How to read what comes back: right, wrong, missing

Every line in those answers falls into one of three buckets. Sort them deliberately, because each bucket means something different and points at a different next step.

**Right** is anything that matches reality as it is today. Current name, current location, the services you actually offer, an accurate one-line description of what you do. Note it and move on. Right facts are the ones you want to protect and reinforce; they are not the problem.

**Wrong** is a stated fact that is false. This is the dangerous bucket, because the platform says it with the same calm confidence it uses for the true things. An old address it presents as current. A service you discontinued, still listed. A sentence that describes a different company under a similar name as if it were you. A customer reading a confident wrong answer has no way to know it is wrong. They just act on it, call the old number, drive to the old address, or rule you out for not offering something you actually do offer.

**Missing** is what the platform simply does not know. You ask what services you offer and it names two of your six. You ask about your service area and it hedges. Missing is quieter than wrong and easier to live with, but it still costs you: a customer asking which provider does a specific job will not hear your name if the platform never learned you do that job.

::::comparison{title="The same business, two platforms"}
:::side{label="Platform A: mostly right"}
Current name. Correct town and service area. Lists the real services. The buyer question names the business as one option. This platform read sources that agree with reality, so its answer is usable.
:::
:::side{label="Platform B: stale and confused"}
Old address from a listing never updated. A service the owner dropped last year, still named. Blends in a detail from a similarly named company. Confident throughout. This platform read a stale source and is repeating it.
:::
::::

There is one more distinction worth learning to hear: the difference between a confident wrong answer and an honest "I do not have enough information." When a platform says it is not certain, or that it cannot find specific details about a smaller business, that is a *missing* signal, and it is far less damaging than a wrong one. A platform that admits it does not know will not mislead a customer. A platform that states the wrong address plainly will. Weight your attention accordingly. The confident wrong answers are the fires; the honest gaps are the empty rooms.

:::callout{type="warn" title="A confident answer is not a correct answer"}
AI platforms state false facts in exactly the same tone they use for true ones. Do not treat any answer as authoritative just because it sounds sure. The entire value of this audit is checking what the platform says against what is actually true, and catching the gap.
:::

## Where a wrong fact most likely came from

This is the payoff, the part that turns a worrying audit into an actionable one. A wrong fact is not the platform making something up out of nothing. It almost always traces back to a real source that the platform read and trusted, a source that is stale, or that disagrees with your own site.

Walk it back. The old address it gave you is probably sitting, right now, on a directory listing or a map profile nobody updated after you moved. The dropped service it keeps naming is likely still on an old profile page, a third-party listing, or a cached version of your own site from before you changed your offerings. The confusion with another company usually comes from a place where both businesses are mentioned with not quite enough detail to tell them apart. When two platforms disagree about you, the wrong one is reading a source the right one did not, and that source is your target.

The reason any of this works is corroboration. An AI platform leans toward facts that several sources agree on, and gets shaky or wrong when its sources contradict each other. That mechanism, why a platform describes a business the way it does and trusts some sources over others, is its own subject; if you want how the recommendation and description actually get assembled, [how AI platforms decide which business to recommend](/blog/ai-recommends) lays out the corroboration logic this audit relies on. For now, the practical move is just to trace each wrong fact back to the most likely stale source feeding it.

You are not fixing anything yet. You are doing the diagnosis: pointing at the listing, the profile, or the cached page that is most likely the origin of each wrong line. That alone puts you ahead of almost every owner, who never gets past "AI is sometimes wrong about us" to "AI is wrong about us *because of that specific outdated listing.*"

:::callout{type="key" title="The audit diagnoses, it does not cure"}
This twenty minutes shows you precisely what is right, wrong, and missing, and points at the likely source of each error. It does not, by itself, correct the record. Seeing the stale listing is not the same as making every source tell the same true story. That correction is the next, separate piece of work.
:::

## Why the wrong facts persist across sources

Once you have traced a few errors back, a pattern usually shows up: the same fact is stated slightly differently in different places, and the differences are what confuse the platforms. One listing has the old address, your site has the new one, a directory has a third version with the suite number dropped. A human glances at all three and understands. A platform reads them more literally, and each disagreement is a small reason for it to hesitate or pick wrong.

That is why consistency across every place you appear, not just a correct homepage, is what actually settles the record. The facts have to match wherever a platform might read them. The underlying idea, naming each fact about your business one exact way and making every source agree, is covered in [keeping your core facts consistent so machines can resolve who you are](/guides/seo/entity-seo-and-eav). The audit you just ran is what tells you which facts are out of sync and where; that consistency work is what brings them back in line.

## What to do with what you found

You now have a labeled document: the right facts (protect them), the wrong ones (each with a suspected stale source), and the missing ones (the gaps to fill). That is a genuine map of how AI sees your business, built from real answers across several platforms, not a guess.

Correcting the record is a real job, and an honest one to name plainly. It means tracking down every stale source, making your core facts identical across all of them, and building enough consistent, corroborated presence that the platforms describe you correctly and start naming you when buyers ask. That delivery, the work of [correcting the record at the source and building the AI visibility once the audit has shown the gaps](/services/aio), is where the findings turn into a fixed story across platforms. The audit is the flashlight; that work is the repair.

Do the twenty minutes first. Open three platforms, ask the five questions, sort the answers into right, wrong, and missing, and trace each wrong fact back to its likely source. Then you will know exactly what needs fixing, and you will be able to tell the difference between someone proposing real corrections and someone selling you a trick. Go run it today, on your own name, on more than one platform, and read what comes back honestly.