Reading a report
A report is the standalone version of a run: one page, no scripts, safe to email and to archive. It reads top to bottom in the order a decision is made.
Masthead
Section titled “Masthead”Which site, when it was analysed, how many pages were examined out of how many were listed, how many templates and machine identities. If pages examined is a small fraction of listed, read the findings as statements about templates, not about every page.
What machines see
Section titled “What machines see”A plain-language account of how each class of machine experiences the site right now, next to four scores: the share of AI-crawler requests that were served, the share of intents fully answered, the share of entities with a defining page, and the finding count after review.
Machine access
Section titled “Machine access”The matrix. Rows are templates; columns are identities; a cell is pages served over pages requested. Anything short of the full count means that identity was refused, challenged or served an error.
A red AI-crawler column means that engine cannot cite the template. A red Plain HTTP column means most retrieval tools see a challenge page. A JS n% chip on a template means that share of its text exists only after JavaScript runs — invisible to crawlers that do not execute it.
Answer coverage
Section titled “Answer coverage”Every intent in the site’s intent space, run through a retrieval engine over the site’s own passages. Answered means a passage fully answers it; partial means a passage exists but is incomplete; unanswered means nothing on the site answers it. The table beneath lists the highest-value questions the site cannot fully answer and what was missing — that is the content roadmap.
Findings
Section titled “Findings”Grouped by severity. Each one names its scope (site, template or page), the pages affected, what to do, what to measure, its evidence, and the knowledge-base hypotheses it rests on with their grades. The count of candidates discarded in review is stated so you know how much was filtered.
What to do, in order
Section titled “What to do, in order”Ranked by impact against effort, with access blockers first — nothing else matters while machines cannot read the page. Each item names the metric and the cohort: the pages to measure on, and the comparable pages left unchanged as a control. The next run reports whether the number moved.
Method and limits
Section titled “Method and limits”What the sample covers, what a user-agent probe can and cannot prove, what the simulation is a model of, and why lab timings are indicative only. Read it before quoting a number to a board.