AI visibility audit

Most companies have no idea what an answer engine says about them. They have rankings, they have traffic, and they have a blind spot exactly where the buying decision now starts. An AI visibility audit closes it. It is the first project in our GEO services, and everything else we plan comes out of it.

What does the audit look at?

Three layers, in the order they break.

Access. Can each engine's crawler reach your pages? We fetch live as each one by name rather than reading your robots file and assuming. Most failures we find are not robots rules at all, they are security presets switched on by someone who never connected them to marketing.

Readability. Once a page is fetched, is there anything to quote? We compare what the raw HTML contains with what a browser shows, check heading structure, check whether an answer exists near the top, and check that your markup matches the visible page instead of contradicting it.

Presence. When a buyer asks the engine directly, are you named? This is the part that decides the plan, and it is the part most audits leave out because it takes real work rather than a crawl.

Why do we test each engine separately?

Because they read different indexes and rarely agree.

About eleven percent of the domains cited by ChatGPT are also cited by Perplexity for the same question. Averaging six engines into one number hides exactly the information you need, which is which engine you are missing and why.

Which crawler belongs to which engine, and what each one tends to cite, is set out in full on the GEO services page.

How do you choose the prompts?

From your sales calls, not from a keyword tool.

We start with the questions your buyers already ask a person: what does this cost, who else does this, is this company any good, what should I ask a supplier before I sign. Then we add category questions in the words a stranger would use, comparison questions naming your competitors, and questions about your own brand by name.

The set is usually 20 to 40 prompts, agreed in writing before we run anything, and run in each language you sell in. We deliver in English and Arabic, and we report the two separately, because the pool of pages an engine can draw on differs by language and so do the answers.

What does the report tell you?

Where you stand, and who is standing there instead.

For every prompt and every engine we log three states: named, linked, or absent. Under each one we list the sources the engine actually used. That list is the plan. If the same review platform appears in nine answers and you have no profile on it, the first task writes itself.

26%

A first audit that comes back empty is normal, not a disaster. It is also the most useful result, because it means the gap is presence rather than something subtle, and presence work has a clear order to it.

How is this different from a free consultation?

Scope and depth.

A free 30-minute consultation is the quick version of the same thinking: we look at one page with you, check crawler rules for the major engines, look for server-rendered text and valid markup, and name the first three fixes. It costs nothing.

The paid audit adds what half an hour cannot: the full prompt set run across several engines in each language, the competing sources named, a read of your third-party profiles and review coverage, a live fetch test against your firewall rather than your robots file, and a ranked plan with an owner and a date on every task.

What happens, step by step?

What does not belong in an audit like this?

An llms.txt check scored as a finding. No answer engine uses the file to decide citations, Google said in June 2026 that it has no effect, and a study of 137,000 domains found 97 percent of them were never requested. We note whether you have one and we do not score it.

A points deduction for missing FAQ markup. Google retired FAQPage rich results in May 2026. Accurate markup that mirrors visible text is still worth having, and the visible questions were always the part that did the work.

A single blended score with no method, a list of 200 findings with no order, and any claim that a particular change will produce a citation. The engines do not work that way and neither do we.

What do you do with the findings?

Fix them in order, or hand them over.

Most engagements move straight into a retainer, where the access fixes ship in the first weeks, the page work runs in weekly batches, and the presence work starts early because it takes longest. The engine-specific detail lives on the ChatGPT optimization and Google AI Overviews and AI Mode pages.

You can equally take the report and run it with your own team. That is a fair outcome and the report is written to survive it: every finding names the page or template it touches and how to test the fix. The rendering and markup items usually land with whoever owns technical SEO on your side. To see how we run this for a team in one of our regional markets, our market page walks through the same process.

How do we start?

Book a free 30-minute consultation first. If we find blocked crawlers or a page with no readable text, you already know the first month of work without paying anyone.

If you want the full picture, book a call and we will scope the audit. Plans are flat monthly retainers with starting prices on the pricing page, month to month, and you can stop at the end of any month.