An AI visibility audit is a one-time measurement of whether your brand appears when AI assistants answer the questions your buyers ask, and of why it does not when it does not. LLMReach runs your category's decision prompts across the assistants, records who gets named and from which source URLs, separates the causes of absence, and hands you a prioritized list of what to fix. It ends with a document and a decision, not with a retainer.

What an AI visibility audit is

An audit answers three questions in order, and the order is the whole method.

Do you appear? For the prompts that sit closest to a purchase decision in your category, is your brand named in the answer, and in what position.

If not, who does? Which competitors occupy the slots you want, and which source URLs the assistant drew from to name them.

Why? This is the part that most measurement skips, and it is the only part that turns a number into work. Absence has causes, they are different from each other, and they cost different amounts to fix.

The output is a document you own: the prompt set, the baseline, the diagnosis, and a ranked list of what to do. You can act on it with us, with your own team, or with somebody else. That is deliberate, and it is worth saying plainly because it changes how you should read the rest of this page.

The problem an audit is built to solve

The real constraint is almost never budget. It is that nobody can say whether the brand appears, and so nobody can say whether anything done about it worked.

Most teams arrive here from one of three places.

They have a suspicion. Someone typed the category question into an assistant, saw three competitors and not their brand, and could not tell whether that was one bad prompt or a pattern. A single query is an anecdote. A prompt set measured on a fixed method is evidence.
They have a dashboard. A tool reports mentions, and the number moves, and nobody can explain why. Movement without a cause is not information you can act on: it is a chart that makes meetings longer.
They have a plan. Content is being written and schema is being added, aimed at AI visibility, with no baseline. Whatever happens next will be unattributable, and six months from now the honest answer to "did it work" will be that nobody knows.

The commercial risk in all three is the same, and it is specific. If a buyer asks an assistant for options in your category and the answer names three companies, being the fourth is functionally the same as being absent. You are not losing a ranking that still gets some clicks. You are being left out of the shortlist before a human evaluates you, and there is no analytics event for it. The audit exists to make that visible, which is the precondition for doing anything about it.

What LLMReach measures, with method

The audit runs in four steps. Each one produces something you keep.

Build the prompt set

LLMReach builds the prompt set with you and you approve it before anything is measured. It is the instrument, and an instrument you did not agree to is a number you cannot argue with later.

The set is deliberately narrow. It covers the questions that sit close to a purchase decision in your category, not every phrasing a person might use. Breadth without prioritization produces a dashboard nobody reads.

Record the baseline

LLMReach runs the set across the assistants and records, for each prompt: whether the brand appears, in what position, which competitors appear instead, and which source URLs the answer drew from.

That last column is the one that does the work later. Knowing you are absent is a number. Knowing which pages the assistant used to answer instead is a map.

Separate the causes

LLMReach separates three causes of absence, because they have different fixes and different costs, and treating them as one problem is the most expensive mistake in this discipline.

A content gap. No page of yours answers the question in a form an assistant can lift cleanly. The fix is content engineering.
An entity gap. The content exists and is fine, but the signals are ambiguous enough that the model cannot confidently attribute the expertise to your organization. The fix is technical and cheap.
An authority gap. Both of the above are healthy, and the answer is assembled from third-party pages where you have no presence. Publishing more of your own content will not move it, and citation graph analysis is what shows where the leverage actually sits.

Rank what to fix

Every gap is ranked by commercial weight and by how tractable it is. You see the reasoning, not just the order. A gap that matters commercially but is expensive to close is a different decision from one that is cheap and marginal, and that decision is yours to make.

What you receive

Deliverable	What it contains	Why it is in the audit
Prompt set	The category questions measured, agreed with you before measurement	It is the instrument every later report is calibrated against
Baseline	Presence, position and cited source URLs per prompt, per assistant	Without a documented starting point, no later claim of improvement can be checked
Competitive picture	Which brands hold the slots you want, and on which sources	Names the specific pages that are answering instead of yours
Cause diagnosis	Each gap classified as content, entity or authority	Decides what kind of work is needed, and what it will cost
Ranked actions	What to fix, in order, with the reasoning shown	Turns a measurement into a plan someone can start on Monday
Technical findings	Entity and structured-data issues found while measuring	These are usually the cheapest wins in the whole document

The boundary matters as much as the list.

Included	Not included
Measurement of prioritized category prompts across the assistants	Exhaustive coverage of every phrasing a buyer might use
Diagnosis separating content, entity and authority causes	Implementation of any of the fixes
A ranked list with the reasoning shown	A guarantee that acting on it produces a position
Findings on entity and structured-data problems	Rebuilding your site platform or CMS
A document you own and can take anywhere	Ongoing measurement after the audit ends

What the audit usually finds

Three patterns come up often enough to be worth naming before you commission anything. None of them is a promise about your case; they are the shapes the diagnosis tends to take.

The brand is present but never first

The assistant names you, in third or fourth position, after competitors. This reads like a small problem and is usually the most tractable one in the document: the content exists and the entity is legible, and what is missing is that no single page answers the decision question directly enough to be the one lifted first. It is a rewrite, not a rebuild.

The brand is absent and the answer is built from third parties

The assistant answers using review sites, listicles and category roundups, and your own pages appear nowhere in the source list. Publishing more on your own domain does not move this, which is why the diagnosis matters: a content program aimed at an authority problem burns a quarter and moves nothing.

The brand is absent and so is everyone else

The assistant answers the category question generically, without naming anyone. This is the finding that most often changes the commercial decision, because the slot is open. It is also the one most likely to close while a team deliberates.

Where an audit fits against the alternatives

Option	What it gives you	What it does not
A single manual query	An anecdote you can show in a meeting	Any way to tell a pattern from a bad prompt
A tracking tool	A continuous number, and movement over time	The cause of the number, or what work would change it
An AI visibility audit	A dated baseline, the causes separated, and ranked actions	Continuity: it measures once and stops
An ongoing engagement	The loop: measure, fix, re-measure	A cheap answer to "is this even a problem for us"

The order that costs least is the one that starts with the cheapest question. If you do not know whether you have a problem, measuring once is a smaller commitment than committing to a program and discovering the answer in month three.

Who it is for, and who it is not for

An audit fits when you need a decision rather than an ongoing program. If the question on the table is "is this a real problem for us, and what would fixing it involve", this is the thing that answers it.

It also fits before committing to anything larger. Starting an ongoing engagement without a baseline means the first month is spent building one anyway, at ongoing rates.

It does not fit two situations, and we would rather say so here than in a call.

You already have a measured baseline and know the gap. Then you do not need another diagnosis, you need execution, and enterprise AI visibility tracking describes the ongoing shape of that work.
Nobody internally can act on the result. An audit that lands in an inbox where no one can approve content or ship a technical change is a document, not a change. That is worth checking before you commission it.

What it costs you to wait

The audit is a measurement, so the case for doing it now rather than next quarter has to rest on what changes in the meantime.

Two things do. The first is that the answer landscape hardens: once an assistant is reliably assembling an answer from a particular set of pages, displacing those pages is more work than being in the set early. The second is internal: every month without a baseline is a month of work you will not be able to attribute afterwards. If your team is already publishing with AI visibility in mind, the measurement is worth more before that work than after it, because after it you can no longer separate what the work did from what would have happened anyway.

Neither of those is an argument for urgency in the abstract. They are the two specific things a delay costs, and if neither applies to your situation, waiting is a defensible call.

Methodology and its limits

Four commitments govern how the audit is run, and each one constrains us.

The prompt set is agreed before measurement. Not after, and not adjusted once the results are in. Choosing the instrument to fit the answer is how measurement becomes marketing.

Every measurement is dated. An assistant's answer is a snapshot. A finding without a date is a claim about the present tense that ages badly and cannot be reproduced.

Absence is reported as absence. If your brand does not appear for a prompt, that is what the document says. There is no partial credit for being mentioned in passing in an answer that recommends somebody else.

Attribution is stated honestly. Where a cause can be identified, the audit names it. Where the evidence does not support a conclusion, it says the cause is undetermined rather than filling the gap with a story. A document that explains every result has stopped distinguishing evidence from narrative.

What this audit does not promise

We would rather set these limits ourselves than have you find them.

No guaranteed position, and no timeline for one. An audit measures and diagnoses. It does not move anything by itself, and nobody controls what an assistant returns. Any provider offering a guaranteed position is describing a mechanism they do not have.

No claim of complete coverage. The audit measures a prioritized prompt set on a defined method. That is a sample, chosen for decision relevance, not every question every buyer asks. Claiming total coverage of a space with no fixed boundary would be overstating what the method can do.

No implementation. The audit ends with a ranked list. Doing the work is a separate decision, and you can take that list to your own team. If you would rather we execute it, that is the AI visibility agency conversation, and it is a different engagement with a different shape.

No prediction of what assistants will do next. The document reports what was observed on the dates it was measured. It does not forecast platform behavior, because we cannot.

How it starts

Scoping. We agree the category, the competitors you care about, and the shape of the prompt set. This is a conversation, not a form.

Approval of the instrument. You approve the prompt set before anything is measured.

Measurement. The set is run and recorded. Nothing is interpreted yet.

Diagnosis and ranking. Causes are separated, gaps are ranked, and the reasoning is written down next to the ranking.

Walkthrough. We go through the document with you. The goal of that session is that you could defend the findings to someone who was not in the room.

Frequently Asked Questions

What is included in an AI visibility audit?

An agreed prompt set, a recorded baseline of presence and position across assistants with the source URLs each answer drew from, a competitive picture of who holds the slots you want, a diagnosis separating content, entity and authority causes, a ranked list of what to fix with the reasoning shown, and the technical findings surfaced along the way. You keep all of it.

How is this different from a tracking tool?

A tool reports whether you are mentioned. The audit explains why you are not, and what class of work would change it. The distinction is the diagnosis: a number tells you there is a problem, a cause tells you what to do about it. Tools are also continuous, and an audit is a single dated measurement.

Do I have to work with you afterwards?

No, and the document is written so that you do not have to. You keep the prompt set, the baseline and the ranked actions, and your own team or another agency can act on them. We think that is the correct arrangement, and it is a useful question to ask anyone else you are evaluating.

How long does an audit take?

It depends on the size of the prompt set and how quickly the scoping conversation converges. The measurement itself is the fast part; agreeing what to measure is what sets the pace. We give you a date at scoping and we do not move it without telling you why.

What do you need from our team?

Agreement on the category and the named competitors, approval of the prompt set, and someone who can answer questions about your own site during the technical review. That is the whole dependency list.

Can you audit a brand in a non-English market?

Ask us during scoping. Coverage varies by assistant and by language, and rather than answer in general terms we would rather tell you what we can actually measure for your specific market before you commit to anything.

What happens if the audit finds nothing wrong?

It gets written down that way. If your category is not being researched through assistants, or your brand already holds the positions that matter, the honest finding is that this is not where your next investment should go. We would rather deliver that than manufacture a problem.

Is the audit the same as the first month of a retainer?

The measurement overlaps, and that is why starting with an audit is the cheaper order. The difference is what happens next: a GEO retainer turns the baseline into an operating loop, and the audit hands you the baseline and stops.

Next step

Start with the free AI visibility audit if you want to see the shape of the measurement before commissioning the full one. It shows which category prompts return your brand, which return competitors, and which source URLs the answers are drawing from.

If you already know you want the full diagnosis, book a call and we will scope it against your category on the call rather than sending you a template.

AI Visibility Audit: What It Measures and What You Get