The standard
This page is the measurement. Everything the checker reports is one of the fifteen checks below, in the same order and with the same names, so a score can be argued with rather than merely believed.
The three questions
A site that AI answer engines never mention fails for one of three reasons, and they need different work to fix.
| Group | Question | Checks |
|---|---|---|
| A | Can AI reach you? | A1–A5 |
| B | Can AI understand who you are? | B1–B5 |
| C | Will AI choose to cite you? | C1–C5 |
Group A is a gate, not a score. If A1 fails — your robots.txt refuses the AI
crawlers — the site is not "partly invisible". It is invisible, and the other fourteen
checks describe a page nobody is allowed to read. A failed A1 caps the report at
"Invisible" and says so at the top.
Group A — Can AI reach you?
| Code | Check | What passes |
|---|---|---|
| A1 | robots.txt allows the AI crawlers |
The named AI crawlers are not disallowed, and the file is not a blanket Disallow: / |
| A2 | Content is in the HTML, not only after JavaScript | The main text is present in the first response, without a browser running scripts |
| A3 | A sitemap exists | A reachable XML sitemap is declared and its URLs resolve |
| A4 | No dead-link pollution | Sampled internal links resolve rather than 404 |
| A5 | Reachable and HTTPS is valid | The site answers over HTTPS with a certificate that validates for the hostname |
Group B — Can AI understand who you are?
| Code | Check | What passes |
|---|---|---|
| B1 | Title and description are complete | A non-empty <title> and meta description on the page checked |
| B2 | Structured data (JSON-LD) is present | Parseable JSON-LD identifying the organisation or the page |
| B3 | Heading hierarchy is semantic | Exactly one <h1> and headings that step down without skipping levels |
| B4 | An llms.txt is present |
A plain-text summary of the site at /llms.txt |
| B5 | The operator behind the site is verifiable | A real identity: a named operator, contact route, and registration details where the law expects them |
Group C — Will AI choose to cite you?
| Code | Check | What passes |
|---|---|---|
| C1 | The conclusion comes first | The answer appears in the opening lines rather than after the preamble |
| C2 | There are extractable structure blocks | Lists, tables and definitions a passage can be lifted from intact |
| C3 | There is original data or first-hand information | Numbers, measurements or experience that do not appear verbatim anywhere else |
| C4 | There is a visible last-updated date | A machine-readable date showing the page is maintained |
| C5 | There is third-party reference evidence | Cited sources that can be checked |
Grades
| Score | Grade | What it means in practice |
|---|---|---|
| 14–15 | Preferred source | Quotable as-is; the remaining work is maintenance |
| 11–13 | Consistently cited | Reliably retrievable, occasionally passed over |
| 6–10 | Occasionally cited | Retrievable but usually not chosen |
| 0–5 | Invisible | Not reached, not understood, or not trusted |
A score without its group breakdown is not actionable, so the report always shows both.
Version
This is standard v1.2. Check definitions change only when the measurement changes, and every change is recorded on this page with the version it belongs to. A report records the standard version it was scored against, so two reports of the same domain taken months apart can be compared honestly.