Where automation is useful
Automated rules are effective at identifying specific, machine-testable conditions across many pages. They help teams repeat checks, spot regressions and collect consistent evidence.
A useful result still needs its page, viewport, element and test condition recorded so another person can reproduce it.
Where human judgement is required
A rule cannot reliably decide whether alternative text communicates the right meaning, whether a focus order supports the task, or whether instructions make sense in context.
Manual keyboard, zoom, reflow and assistive-technology procedures cover behaviour that a page-level rule cannot determine on its own.
Use evidence, not a headline score
An automated score is a directional signal for the tested scope. It is not a compliance percentage. Delivery decisions should use reviewed findings, coverage and documented limitations together.