The short version: an AI visibility audit answers a question no analytics tool reports — when your buyers ask AI systems for a recommendation, what actually comes back, and why. It's a diagnostic rather than a report card: the point isn't a score, it's establishing which specific layers are keeping you out of the answers, since the four common causes look identical from outside and require completely different work. Here's what the process examines, the kinds of findings that come back, and why they usually surprise people.
What gets tested
The foundation of any honest audit is the probe layer — actually asking the AI engines the questions your buyers ask and recording what they say. Not brand-name queries, which tell you little, but the real commercial questions: who's best for this need, in this category, for this kind of customer. Across the major engines, repeated for consistency, with the answers captured verbatim. This produces the one thing no dashboard gives you: documented evidence of where you stand right now, and who's standing there instead of you.
Around that sit the diagnostic layers that explain the probe results. Entity clarity — can a machine determine unambiguously what your business is, who it serves, and where. Content citability — is there anything on your site specific enough to be quoted in a generated answer, and does it exist in the formats your buyers' question types call for. Corroboration — what independent sources say you exist and are credible, and how that compares to the businesses being named instead of you. Index presence — whether your pages are actually present in the underlying indexes the major assistants retrieve from, which is checked rather than assumed because it's frequently the surprise.
The findings that come back most often
"We're not mentioned, and here's who is." The headline finding for most businesses, and the one that makes the problem concrete. It's rarely a surprise that you're absent — most owners suspect it — but seeing exactly which competitors are being named, for which questions, tends to land differently than suspecting it.
"The engine is citing sources you've never heard of." Consistently the most useful finding. When an AI justifies a recommendation, the sources it leans on reveal where authority in your category actually lives, and it's usually not where businesses assume. It's frequently not the competitors' own websites at all but third-party content — roundups, directories, review platforms — that most businesses have never considered part of their marketing. This finding alone often reframes a company's entire strategy.
"Your identity is ambiguous or wrong." When probing turns up an engine describing a business inaccurately, conflating it with another company, or unable to characterize it at all, that's an entity problem with knock-on effects everywhere else. Businesses are usually startled by this one because their website seems perfectly clear to them — the ambiguity is only visible from the machine's side.
"You're missing from an index you assumed you were in." The most mechanically addressable finding and among the more common. Everything appears healthy from the business's perspective, so nobody thinks to check, and the absence quietly caps visibility on one or more major assistants regardless of everything else being right.
"You have nothing quotable." Often the finding that stings, because it's not about anything being broken — the site works, it's well-designed, the company genuinely has expertise. It just hasn't written any of that expertise down in a form a machine can lift into an answer, so the engine has nothing to work with and cites someone who did.
Why the results usually surprise people
Three things account for most of the surprise. First, businesses generally expect their website to be the problem and are startled by how much of the gap sits off their domain in the corroboration layer — a place they don't think of as marketing and can't directly edit. Second, the competitors being named are frequently not the ones they consider their real competition; engines often surface businesses that invested in being legible and corroborated rather than the strongest operators in the market. Third, the causes rarely match the assumption: companies expecting a technical finding often have a content problem, and companies expecting a content problem often turn out to be missing from an index.
That last pattern is the practical argument for auditing before acting. The four main causes produce an identical outward symptom — you're not in the answers — while requiring entirely different work on entirely different timelines. Index presence can move in weeks. Content citability is a matter of months. Corroboration compounds over quarters. A business that guesses wrong doesn't just waste the effort; it concludes that AI visibility work "doesn't do anything," when in fact it addressed a layer that wasn't the constraint.
What you get from a diagnosis
The output that matters isn't a score — a number expressing your visibility is easy to produce and almost useless, because it doesn't tell you what to do. What's actually useful is: the verbatim answers the engines gave to your real buyer questions, so you can see the problem yourself rather than take anyone's word for it; the specific sources being cited instead of you, which is effectively a map of where the authority in your category sits; a determination of which layers are failing and in what order they'd need to be addressed; and an honest read on timeline, since some of this moves quickly and some genuinely doesn't.
That's the difference between a diagnostic and a sales instrument. A good audit should be useful to you even if you never engage anyone to act on it — you'd know where you stand, who's winning, why, and what sequence the work would follow. That's also why we credit the cost of our audit toward implementation if you decide to proceed: the diagnosis has standalone value, and it shouldn't be a toll gate. What it buys you is the ability to act on evidence rather than on assumption, which in a discipline this new is the difference between progress and expensive guessing.
FAQ
What is an AI visibility audit?
A diagnostic that establishes whether AI systems recommend your business when your buyers ask, and if not, why. It probes the major engines with your real commercial questions, records the answers verbatim, and then examines the underlying layers — entity clarity, content citability, third-party corroboration, and index presence — to determine which are actually failing.
How is this different from an SEO audit?
An SEO audit examines how well your site is positioned to rank in classic search results. A visibility audit asks a different question: when an AI composes an answer, are you named and cited? That adds layers traditional SEO doesn't assess — whether machines can resolve your identity, whether your content is quotable, whether independent sources corroborate you, and whether you're in the specific indexes AI assistants retrieve from.
What's the most common finding?
That the sources an engine cites to justify recommending someone else are third-party content the business never considered part of its marketing — roundups, directories, review platforms — rather than competitors' own websites. It's the finding that most often reframes strategy, because it reveals that a meaningful part of the gap sits off your domain entirely.
Will an audit tell me my business is doing badly?
It tells you where you stand with AI systems, which correlates poorly with how good your business is. Excellent companies are frequently invisible while more ordinary competitors get named, usually because the latter are better corroborated rather than better at the work. The finding isn't a judgment on your business; it's a map of a specific, addressable gap.
Is the audit useful if I don't hire anyone to fix it?
It should be. A proper diagnostic leaves you knowing where you stand, who's winning your questions, which sources hold the authority in your category, which layers are failing, and roughly what timeline the work would run on. That's genuinely actionable regardless of who acts on it — and it's why we credit the audit cost toward implementation rather than treating it as a gate.
How long does an audit take?
A thorough one is a matter of a couple of weeks rather than days, because it involves probing multiple engines across a set of real buyer questions, repeating for consistency, and then examining each underlying layer to establish cause rather than just symptom. Something delivered in an hour is almost certainly an automated score rather than a genuine diagnosis.
Can't I just check this myself?
You can absolutely establish the symptom yourself — ask the engines your buyer questions and see whether you're named. That's worth doing and costs nothing. What's harder is determining which of the underlying layers is responsible, since they present identically from outside, and getting that wrong means directing effort at a layer that isn't your constraint. The symptom is easy to observe; the cause is the part that takes real diagnostic work.
Find out exactly where you stand. Our AI Visibility Audit is $2,500, credited toward implementation, and delivers the transcripts and the gap map. Related: how to tell if AI can see your website and our own GEO case study.