Conformance

A clean scan is not, on its own, a WCAG conformance claim — and pretending otherwise is exactly the kind of overreach this engine refuses to make. This section covers how individual rule results actually roll up to a Success Criterion verdict, what that verdict honestly does and doesn't cover, and how to gate a real CI pipeline on known, already-reviewed violations instead of blocking every build until everything is fixed at once.

WCAG conformance mapping

See exactly how atomic rules, coverage facets, and composite WCAG-SC rollup rules fit together, and why a criterion whose checks all came back clean reads cantTell, partly checked, rather than a vague "looks fine": a composite says which parts were checked and leaves the rest to a person. A scan that names a few checks rather than a target still reports the criteria they belong to, and says which ones it only partly checked. Target a conformance level (A, AA, or AAA) across any WCAG version with one tag set or a named profile, map every result to the EN 301 549 clause that restates its criterion when that's the standard you report against, and get a straight answer to the question that actually matters: what can this kind of automated result tell you, and where does it stop. See WCAG conformance mapping.

Test procedures

A scan is half an assessment; the other half is a person checking what rules can't decide. Core gives that half too: for every criterion its rules evaluate, a procedure for each facet, with the steps, the tools it needs and what passes and what fails, in five languages. Read them as data, or as a reference page or a worksheet that records each result and rolls them up per criterion, with no scan or next to one, so an audit and a CI run work from the same text. See Test procedures.

Baseline / allowlist

Don't let a first scan of an imperfect, existing site block every future build — record what's already known once, and gate CI only on violations that are genuinely new. It's the difference between "we can't ship anything until this is perfect" and "we can catch regressions starting today," and it's what actually gets a team to turn the gate on. It can accept the findings that likely fail too, one baseline covers a scan of several profiles, and a baseline written before an upgrade keeps matching for one minor release when a finding's reason code is split or newly given. See Baseline / allowlist.