How to Audit Your Brand's Visibility in AI Search

Most people who understand AI search in theory have never tested their own brand in it.

They know the tools matter. They've read about citations and topical authority. But they've never opened ChatGPT, typed their own company name, and looked hard at what came back.

That's worth fixing, because AI visibility is measurable. Not precisely — there's no Search Console for ChatGPT — but well enough to tell whether you're invisible, whether you're visible but described wrongly, and whether you're being beaten by a specific competitor. Those are three different problems with three different fixes, and you can't tell them apart without looking.

This is the brand audit for AI search that I run. It takes about an hour, and it's repeatable, which matters more than any single result.

---

First, the thing that makes AI audits different

Search Console gives you one number because Google gives everyone roughly the same results page. AI tools don't work that way.

Two things break the comparison, and you need to account for both or your audit will mislead you:

Answers vary between runs. Ask the same question twice and you can get different wording, different examples, sometimes a different set of businesses. One absence isn't evidence. A pattern across several runs is.

Your account has history. If you've spent months talking to ChatGPT about your own business, it may have memory or personalisation working in your favour in a way no prospect shares. Run every audit query in a logged-out session, a temporary chat, or a private window. Otherwise you're measuring your own footprint, not your visibility.

That second one invalidates more DIY audits than anything else. It's also the easiest to get right.

---

What you're actually testing

Three questions, in increasing order of commercial value:

1. Does AI know I exist? — brand queries
2. Does it describe me accurately? — accuracy queries
3. Does it recommend me for what I do? — category and problem queries

Most businesses pass the first, fail the third, and have never checked the second. The third is where the money is, because that's the query a prospect runs when they don't yet know your name.

---

Step 1: Set up the log before you start

Open a spreadsheet first. If you run 30 queries across four platforms and try to remember the results, you'll remember the annoying ones and nothing else.

Columns: query, platform, date, mentioned? (yes/no), position in answer (first, middle, buried), described accurately?, sources cited, who else appeared.

That last column is the one people skip and the one that pays. Your competitors' names appearing consistently is your list of who to study.

---

Step 2: Brand queries — does AI know you exist?

Logged out, run these on each platform:

• "What is [your brand name]?"
• "Tell me about [your company]"
• "Who is [your name]?"
• "What does [your brand] do?"

You're checking three things: whether it knows you at all, whether the description is right, and what it cites as its source. That last one is the most useful output of the whole step. If it's citing your homepage, good. If it's citing a directory listing, a five-year-old press mention or your LinkedIn, that tells you what the model considers authoritative about you — and it's often not your website.

If it confidently invents something — a wrong location, a service you don't offer, a founding date out of nowhere — write the exact wording down. That's a correction target, and it usually traces back to thin or conflicting information about you on the open web.

---

Step 3: Accuracy queries — is what it says actually right?

Push past the summary:

• "What services does [your brand] offer?"
• "Who does [your brand] work with?"
• "Where is [your brand] based?"
• "Is [your brand] any good?"

The last one is uncomfortable and worth doing. It surfaces what review and reputation signals the model has picked up, which is often nothing — and "I don't have information about their reputation" is itself a finding that points straight at your review profile.

---

Step 4: Category and problem queries — the ones that matter

This is the real test, and where most audits get interesting.

Search the way a customer would — describing the problem, not your job title. Customers don't search for "generative engine optimisation consultant." They search for "my website isn't getting enquiries."

• "Who are the best [service] specialists in [your area]?"
• "I need help with [problem you solve] — who should I talk to?"
• "Best [category] for a small business in 2026"
• "How do I find someone to [do the thing you do]?"

Run each at least three times, in fresh sessions. Note who appears and how consistently. If the same three competitors come back every time and you never do, that's not bad luck — that's a topical authority gap with names attached.

---

Step 5: Comparison queries

Closest to the decision point:

• "[Your brand] vs [competitor]"
• "Should I use [your brand] or [alternative]?"
• "Alternatives to [competitor] for [use case]"

Even total absence here is useful. The audit tells you who is being named at the comparison stage, and you can then go and look at what those businesses publish, where they're cited, and what they make easy to quote.

---

Step 6: Run it across platforms that work differently

Don't audit ChatGPT alone and call it AI visibility. The tools draw on different sources, and the differences are diagnostic:

• ChatGPT — mix of training data and live browsing depending on the query
• Perplexity — heavily citation-led, usually shows its sources plainly
• Google AI Overviews — closest to classic search signals, so it tends to track your existing rankings
• Gemini and Copilot — different source pools again

The pattern tells you the cause. Present in Perplexity and AI Overviews but absent from ChatGPT usually means your content is findable and citable, but you're not established enough in training data — a slower, mention-driven problem. Absent everywhere including Perplexity usually means the raw material isn't there to cite, which is a content and structure problem you can start fixing this week.

---

---

Step 7: Read the patterns, not the individual answers

With the log filled in, look across it rather than at any one row:

What you see	What it usually means
Visible on brand queries, absent from category queries	Topical authority gap — you're known but not associated with the subject
Mentioned but described wrongly	Conflicting or thin information about you across the web
Visible in one platform only	Source-pool difference; check what that platform cites
Competitors named consistently, you never	They have citable third-party coverage you don't
Nothing anywhere, including your own name	Entity problem — nothing authoritative establishes you exist

That last row is the most common outcome for small businesses, and the most fixable. It's rarely about the writing. It's that nothing outside your own website corroborates you.

---

What to do with what you find

Three broad fixes, matched to the three diagnoses above. Each has a full walkthrough elsewhere, so briefly:

If nothing can cite you — the problem is content that answers questions directly. AI tools surface sources that state things plainly and early. Burying the answer under 400 words of preamble makes a page hard to quote. Getting your business to show up in ChatGPT covers the structural side of this.

If you're described inaccurately — the web disagrees about you. Your name, description and positioning need to match across your site, LinkedIn, any directory listings and any press. Conflicting information doesn't just get ignored; it gets averaged into something wrong.

If competitors are named and you aren't — this is the Authoritativeness problem, and it's the hardest and slowest. It's earned through third-party mentions from sources that already carry weight, which is unpacked in E-E-A-T explained.

Structured data is worth having under all three. It won't invent authority you haven't earned, but it makes what you do have unambiguous — and if you're not sure what AI search is optimising for in the first place, how to get your business to show up in ChatGPT covers the picture from the other side.

---

Make it quarterly

One audit is a snapshot and tells you very little. The same audit run four times a year tells you whether you're gaining ground.

Keep the query list fixed — don't improve the wording between runs, or you lose the comparison. Same queries, same platforms, logged out, once a quarter. Track the percentage of category queries where you appear at all, and watch that one number over a year.

Everything in AI search is moving. Your position in it is only meaningful as a trend.

---

FAQ

How do I check if my business appears in ChatGPT?

Open a logged-out or temporary chat and ask about your business directly — "What is [brand]?" — then ask category questions a customer would use, like "who are the best [service] specialists in [area]?". Run each several times, since answers vary between sessions. Staying logged out matters: a personalised account can surface your own business because of your history, not because a prospect would see it.

Why does ChatGPT recommend my competitors instead of me?

Almost always because they have third-party coverage you don't. AI tools lean on sources that corroborate a business independently — directories, reviews, press, industry sites. If everything written about you lives on your own website, there's little for a model to draw on. It's rarely about the quality of your writing.

Is there a tool that tracks AI search visibility?

Paid tools exist and are improving, but a manual quarterly audit with a fixed query list and a spreadsheet gets a small business most of the value. The important discipline isn't automation — it's running the same queries, logged out, and comparing over time.

Why do I get different answers each time I ask?

These systems are probabilistic and often browse live, so wording and examples shift between runs. That's why one absence isn't a finding. Run each query at least three times and record how consistently you appear, rather than treating any single answer as the result.

How often should I audit my AI search visibility?

Quarterly suits most businesses. Keep the query list identical between runs so results stay comparable — the trend across a year is far more informative than any single audit, and rewriting your queries each time destroys the comparison.

Does structured data help with AI search visibility?

It helps by removing ambiguity about who you are, what you do and how to reach you, which makes you easier to summarise correctly. What it won't do is manufacture authority — schema describes your business accurately, but being recommended still depends on other credible sources referencing you.