The Policy Health Score ranges from 0 to 100 when sufficient evidence is available. It is not a compliance score, a legal score, a security score, or a protection score — it is a weighted summary of how well a website's declared AI crawler policy matches its selected objective.
Weighting
| Category | Weight |
|---|---|
| Resource availability and validity | 15% |
| Syntax and deterministic evaluation | 20% |
| Business-objective alignment | 30% |
| Cross-signal consistency | 15% |
| Registry freshness and explicitness | 10% |
| Monitoring and change risk | 10% |
Labels
| Score | Label |
|---|---|
| 90–100 | Strong |
| 75–89 | Good |
| 50–74 | Needs attention |
| 25–49 | Weak |
| 0–24 | Critical |
| — | Incomplete (insufficient evidence) |
Optional declarations are neutral
Since ruleset 2026.10.1 (15 October 2026), publishing an optional declaration — an RSL licence, Content Signals, llms.txt or a sitemap — neither adds to nor subtracts from the score. Only a declaration that is present but broken costs points: a licence that cannot be retrieved or does not conform to RSL 1.0, a malformed or ambiguous Content-Signal line, or a sitemap that is not a sitemap. Informational robots.txt notes, such as a recognised extension field, are not syntax problems. The score never rewards adding a directive you don't need. Scans evaluated under ruleset 2026.07.2 keep the scores they were given.
When there is not enough evidence
When a scan cannot retrieve enough policy resources to evaluate confidently, CrawlPact shows "Incomplete" rather than a misleading number — never a gauge that implies false precision.
Reproducibility and comparisons
The same declared policy, evaluated against the same registry version and ruleset version, always produces the same score — see methodology for why. This means a score can change between two scans of an unchanged website for a reason that has nothing to do with that website: a registry or ruleset update. CrawlPact labels a registry update "registry drift" and distinguishes it from "website drift" (a change in what the site itself declares); every report and finding also names the ruleset version it was evaluated with — comparing scores across a registry update is comparing against a different, newer yardstick, not evidence the site got better or worse.