AI visibility audit
What gets examined, what comes back in the report, and the part most audits in this category quietly leave out. You can run the mechanical half yourself, free, before deciding whether to buy anything.
An AI visibility audit measures whether AI assistants can reach, read and quote your site, and whether they currently name you. It has two halves that are usually sold as one: a deterministic technical half anyone can verify on your live pages, and a sampled half that varies between runs and has to be reported as a rate.
- The two halves
- Deterministic: crawler access, rendering, schema, freshness. Sampled: whether assistants name you.
- What is verifiable
- The technical half. Any third party can re-run it on your live pages and get the same answer.
- What is not
- The sampled half. Generative answers vary between runs, so it is a rate, never a screenshot.
- Cost to start
- Free. The technical half runs on any URL with no card.
The two halves, and why the split matters more than the score
Almost every audit sold in this category blends two very different kinds of measurement into a single number, and the blend is where the value leaks out. Separating them is the first thing a good audit does, because the two halves have different reliability, different costs and different fixes.
The deterministic half is everything about your own pages: which crawlers are allowed, what comes back before JavaScript runs, whether the structured data parses, whether the page carries a date a machine can read, whether the answer sits in a passage that can be lifted. Every one of these is a fact about a URL at a moment. Run it twice and you get the same answer. Hand it to somebody else and they can check it.
The sampled half is whether assistants actually name you. This is a real measurement and it is not deterministic: ask the same question twice and the answer can differ. It is honest only as a rate across a fixed set of questions, per engine, rerun on a cadence.
The reason to insist on the split is money. The deterministic half is cheap, fast and fixable, and it is where most problems actually are. The sampled half is the one people buy first and it is the one that tells you least about what to do next. An audit that leads with a single blended score is selling you the expensive half of the answer.
What is actually in the report
This is the question behind most searches for an AI visibility report, so here is the contents page rather than an adjective. Our own report is scored per page across the whole crawl, and every finding names the URL it was found on.
Per bot crawler access
Each named agent listed separately, allowed or refused, read from your live robots.txt with the real matching rules applied. Training crawlers and retrieval crawlers are shown apart, because blocking one does nothing about the other and confusing them is the most expensive mistake in this field.
What a crawler without JavaScript receives
The rendered and unrendered versions compared. A page whose content only appears after the client runs is a page an assistant receives empty, and it looks like a content problem from outside.
Indexability and snippet controls
Whether anything on the page quietly tells engines to skip it or not to quote it. A noindex, a nosnippet or a data-nosnippet around the passage worth quoting removes you from the surface regardless of everything else.
Structured data, and whether it matches the page
Not just whether the markup parses, but whether it describes what a reader actually sees. Schema that contradicts the visible page is worse than none.
Quotability, per passage
Whether each section answers one question in a self contained block, whether figures sit next to the things they describe, and whether the page carries a date and a named author. This is the half that decides whether retrieval turns into a citation.
The prompt results, per engine
Your buying questions put to the assistants, recording whether you were named, whether you were linked, who was named instead, and which sources the answer drew on. The competitor list is reliably the most read line in the whole report.
Every finding with the URL and the fix
A finding without an address is an opinion. Each one carries the page it was found on, why it matters and what to change, so the report can be handed to whoever does the work without a meeting in between.
What an audit cannot tell you
Three things, and any audit that claims otherwise is overselling.
It cannot tell you what it will be worth. Assistants pass very little referrer data, so traffic that began in an AI answer usually arrives looking like direct. Anybody quoting a revenue figure attributed to AI visibility is modelling rather than measuring, and the honest version of that claim names its assumptions out loud.
It cannot promise placement. There is no position to buy and no schema that guarantees a citation. There is eligibility, which is technical and checkable, and there is being the best available source for a question you never see, which is editorial and slow.
It cannot fix the half you do not own. Most of what an assistant says about a company comes from pages that company does not control. An audit can show you which third party sources an answer drew on, and it cannot get you onto them. That work is people, not software.
What it costs
Every limit below is enforced by the server rather than advertised loosely, and the free tier is a real tier rather than a trial.
Nothing, for the deterministic half. The technical audit runs on any URL with a free account and no card, and the report is yours to keep whether or not we ever speak. There is no trial timer on it and no page limit worth mentioning at this size.
Paid plans exist for the things a one off audit cannot do: more pages per crawl, prompt tracking across engines, and history so you can see direction rather than a single reading. Those start at $24 a month and the ladder is set out on the pricing page. If you want the audit run and the fixes shipped by our team rather than handed to yours, that is a service plan and it starts separately at $99 a month.
Questions people ask about AI visibility audits
What is an AI visibility audit?
A measurement of whether AI assistants can reach, read and quote your site, and whether they currently name you when somebody asks a question you should be the answer to. It has a deterministic technical half that anyone can verify on your live pages, and a sampled half that varies between runs and has to be reported as a rate rather than a screenshot.
How much does an AI visibility audit cost?
The technical half is free here, on any URL, with a free account and no card. Paid plans add prompt tracking across engines, more pages per crawl and history. If a vendor quotes four figures for a one off audit, ask which parts of it are deterministic, because that is the part you can get for nothing and check yourself.
How long does an AI visibility audit take?
The technical crawl is minutes rather than days. What takes time is the sampled half, because a single run is a single sample and a defensible rate needs the same prompt set asked repeatedly. Any audit that hands you a confident visibility percentage from one afternoon of asking has measured one roll of a die.
What is in an AI visibility report?
Per bot crawler access, what a crawler without JavaScript receives, indexability and snippet controls, structured data validity, quotability per passage, and the prompt results per engine with the competitors named in your place. Every finding carries the URL it was found on and what to change, so it can be handed straight to whoever does the work.
Can I do an AI visibility audit myself?
The deterministic half, yes, and you should before paying anybody. Crawler access, render output, indexability and structured data are all checkable free on any URL in about a minute each. The sampled half is harder to do well by hand, not because asking is hard but because doing it repeatedly and consistently enough to be a rate is tedious.
How often should it be repeated?
The technical half whenever the site changes materially, since it is cheap and it is where regressions hide. The sampled half on a fixed cadence, because the only useful version of that number is a trend. A one off reading with nothing to compare it to tells you where you are and not whether you are moving.
Run the deterministic half now, for nothing
Crawler access, render output, indexability, structured data and quotability, scored across your site, with every finding attached to the URL it was found on. Free, no card, and the report is yours whether or not we ever speak.