Does each page answer what it is shown for?
What it is
On each project page, the card "Does each page answer what it is shown for?" grades the pages we published to your own site. Two or three AI models from different companies, with web search off, are each asked to quote the passage that answers the page's question. Every quote is checked against the page's own text, so a model cannot pass a page by agreeing with it.
Two kinds of result, never added together
OBSERVED-QUESTION ASSESSMENT — Search Console reported a real search that showed your page. That search is the question. It carries a verdict (Answers, Does not answer, or Contested when the models disagree), how many graders stood behind it, and the Search Console row: how often it was shown and at what position. TOPIC DIAGNOSTIC — there is no such search yet, so the page is asked about its own subject. It carries NO verdict, because a page is written to answer its own subject and passing that says little. It carries what is useful: the lines a writer could quote (checked against the page), the kind of source the models think the question expects, and what they say is missing. The headline counts pages analysed and splits them. There is no percentage, on purpose.
What the thumbs-up means
A thumbs-up on a diagnostic means the models' majority judged the page the right kind of source for its topic. It is their opinion and is labelled as one. It does not mean the page "passed" — a diagnostic has no verdict to pass. Whether asking the shape questions in the same call moves a grader's verdict was tested on 2 September 2026: 37 documents, three graders, each document graded with the questions, without them, and with them again as the noise floor. No verdict flipped (0 of 30, 0 of 31, 0 of 35). The test is repeated whenever the prompt changes.
Why a page has no search yet
The card says why, because the reasons need different fixes. "Google reported no search for this page in the last 90 days" is about the page. "Search Console is not connected", "no property covers this domain" and "we could not read Search Console" are about the connection, and only connecting it changes them. Search Console reports up to three days ago, so a page published this week always reads as "no search yet".
Rebuilding a page in the right shape
Where the models say a page is not the right shape, "Rebuild in the right shape" hands their brief to your default AI writer with the page's own text as the only source. The rewrite keeps the verified lines word for word, may add no facts, and may link only sources already on the page or recorded against your business in the Entity Factory — never an invented reference. On a connected WordPress site the post is updated in place at the same address. Elsewhere the rebuilt page is held on the card for you to copy across. Either way it is graded again on the next nightly sweep, so before and after are two readings.
What it costs
Grading runs on your own AI keys: two or three calls per page, once, nightly, newest first, up to ten pages a night. A page is graded again only when the question changes, our text extractor changes, or you rebuild it. The weekly check that a verified passage is still on the page reads the page once and uses no AI. A rebuild is one call to your default writer.
Answerability says nothing about whether an engine will surface the page. Visibility is measured separately, per engine, and the two are never averaged.