Buyers are already asking assistants for recommendations before they ever reach a website. Most businesses have no idea what they are allowing those systems to do, and most vendors selling a fix cannot tell them either.
GPTBot, ClaudeBot, PerplexityBot, Google-Extended and CCBot are each named separately, and a well-meaning blanket rule can block all of them at once. Most site owners have never read the file. Fewer have read what it says to each agent individually.
Tags injected at runtime appear in no fetched HTML. That makes "is anything tracking this" genuinely unanswerable from outside — which is why we say so rather than scoring you down for it, and turn it into the question actually worth asking.
An assistant cites what it can parse: server-rendered text, structured data, clear headings, a resolvable canonical. A site that renders its copy in JavaScript can look fine to a human and be invisible to a model.
llms.txt and the .well-known endpoints exist precisely so a site can state its own identity and citation preferences to AI systems. Published wrong, they attribute your content to somebody else.
None of that requires a guess. All of it is observable from outside your site in an afternoon — which is exactly what this is.
01
What your robots.txt says to each AI crawler by name, not just to search engines in general. GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot and anything else declaring itself.
Whether a Content-Signal is published, and what it asserts. Whether llms.txt exists and whether it names you correctly.
We honour Disallow during the audit itself, including where we could technically route around it. A vendor who ignores the one machine-readable place you can say no is telling you something about how they will treat your site later.
02
Whether your content is server-rendered and therefore parseable, or assembled in the browser and therefore not.
What structured data exists and what types. Whether canonical resolves, and to where. Whether headings carry meaning or styling.
Whether there is anything on the page an assistant could lift as an answer and attribute to you.
03
The primary call to action, quoted verbatim as it appears. The length of your main form, counted.
Where a reading cannot be made — a JavaScript-rendered form has no fields in fetched HTML, and zero fields is not the same as a frictionless form — we report it as unread rather than as a compliment you did not earn.
Copy findings carry a quote from your own page. A finding we cannot quote is dropped before you see it.
04
Every section reports what it could not assess and why.
A missing finding otherwise means either your site is fine or we never looked, and from the outside those are identical. That ambiguity is where every padded audit hides.
You get the raw signals we read, with the date we read them, so you can check our work or re-run it yourself later.
If you have had a free AI-readiness scan in your inbox this year, you have probably seen all four of these.
One fixed fee. Five business days. The whole fee comes back to you if you go further.
If the audit finds little worth fixing, we will tell you that, and the engagement ends there. A diagnostic that can only produce one answer is not a diagnostic.
The free scan takes about two minutes and gives you the headline findings. The paid audit is what quotes every one of them.
Because a free scan has to produce an alarming result to earn the next call. Ours is paid, which means it is allowed to tell you that your robots.txt is fine, your schema is adequate and the real problem is somewhere we were not hired to look. That option is what you are paying for.
No, and nobody reading your site from outside can. Appearing in an AI answer is measured by querying those systems repeatedly over time against a fixed prompt set, which is a different piece of work. This audit reads what your site permits and publishes. We separate the two deliberately because conflating them is how the whole category misleads people.
Every finding quotes the thing it looked at. A rule that cannot name the specific directive, headline or CTA it is judging does not fire at all, and a copy finding whose quote we cannot locate verbatim in your own page is dropped before delivery. That is enforced in the code, not promised in the pitch.
No. We read robots.txt first and honour Disallow, including in the cases where we have the means to get through anyway. If a vendor will route around the one machine-readable place you can refuse automated access, consider what else they will treat as optional.
It comes off the first invoice in full, provided you engage within 30 days of delivery. The audit is how we find out whether there is work worth doing, not a profit centre.
Yes. The signals we read are part of the deliverable, with the date attached. An audit you cannot check is an opinion, and an observation with no date on it goes stale without telling you.
Start with the free scan — it costs you two minutes and tells you whether there is anything here worth $2,500. If there is not, we would rather you knew that now.