Can an AI assistant quote this page?
How the score is calculated →·Compare against rivals →·Find content gaps →·Read your access log →
Most "AI visibility" checkers count crawlers. This one asks what blocking each crawler costs you — because blocking a training bot costs nothing and blocking a search bot costs the citation — and then checks whether there is anything on the page worth quoting once they arrive.
What it checks
- Retrievability. Thirty-six crawlers against your robots.txt, each one labelled with what it does — search, fetch-on-request, or training. Blocking the third costs you nothing and we say so.
- Readability. Whether the served HTML carries the answer, or whether a browser has to build it. Assistants do not run your JavaScript.
- Answerability. Whether a passage lifted out of the page still makes sense to somebody who never read the sentence before it.
- Structure and discovery. Who the page says publishes it, and how a crawler is meant to find your other pages.
Anything we cannot measure is reported as unmeasured rather than scored as a zero. A site that refuses our scanner has not failed; we just do not know, and a tool that turns that into a number is guessing at you.
What answer engine optimisation actually means
Search sends you a visitor who then reads your page. An assistant reads your page and sends the visitor an answer. The page still has to be fetched, but it now has to survive being summarised by something that will quote one paragraph of it to somebody who never sees the rest — and the things that make a page rank are not the things that make a paragraph survive that.
So there are three questions, in order, and the second and third only matter if the first is yes. Can it be fetched — 36 crawlers across 21 vendors, each of which your robots.txt either permits or does not. Is there anything to read when it is — assistants fetch HTML and do not run JavaScript, so a page a browser assembles at runtime arrives empty however open robots.txt is. Does a passage stand on its own — a paragraph beginning “It costs nothing” is useless when quoted alone, and that is how it will be quoted.
Blocking a training crawler costs you nothing
This is where most advice in this area goes wrong. Of the 36 crawlers we check, 16 exist to collect training data, 12 to answer a question with a citation, and 8 to open a link somebody has already named. Blocking the first group has no effect on whether you are cited — it is a publisher's decision about their own work, and plenty of good ones make it. Blocking the second costs you the citation outright.
A tool that counts “AI bots blocked” as one number tells you nothing, because the number mixes a choice with a mistake. We label every crawler with what blocking it actually costs, and score a blocked training crawler at zero rather than marking a decision down as a defect.
Why a page can rank well and never be quoted
An assistant retrieves passages, not pages. A page split into chunks by somebody else's retriever can have its heading land in one piece and the answer to that heading in the next, so the passage arrives without the question it answers. The same page scores perfectly in every SEO tool you own.
Under 250 words of served HTML we stop scoring readability rather than guessing at it — a verdict carrying fifty per cent confidence is a coin toss, and turning a coin toss into a number is what the confidence figure exists to prevent. That threshold caught this page, which is why these paragraphs are here: it served 225 words and could not be judged.
Every weight, threshold and refusal is published, read out of the code that scores on it so it cannot drift. The methodology page lists them, including the five things we measure, show you, and refuse to put a number on.