Text without enough contrast.

What fails — and who it fails

The failure, in plain language.

Body text and interactive labels need 4.5:1 against their background (3:1 for large text). Brand-light greys and tinted buttons routinely land at 2–3:1 — legible to a designer on a calibrated screen, gone for everyone else.

Detection — what the machine can and cannot judge

How the scanner finds it.

• Foreground and background colors are read from the rendered page — computed styles, not the stylesheet — and the exact ratio is reported per element.
• Text over gradients or images is marked for review rather than guessed.
• What automation cannot judge: brand trade-offs. The fix is usually one design token, and the finding shows exactly which one.

The legal reading — factual, not fearful

Where this sits in enforcement.

1.4.3 is Level AA and one of the most frequently evidenced criteria in assessor reports: the ratio is a number, and a number is an exhibit. It is also the cheapest large-scale fix on most sites — a token change clears every instance at once.

The fix — by pattern

2 patterns cover nearly every instance.

Fix the template, not the page — one change typically clears every instance at once. The triage engine groups findings by template pattern for exactly that reason.

Verification — resolution is evidence, not a checkbox

Fixed means a scan said so.

On monitored domains, mark the finding fixed and the next scan verifies it: zero remaining instances converts the finding to Resolved, and both the claim and the confirmation land in the append-only ledger. Criteria that need human judgement close through sign-off in the coverage matrix — covered by the right method, never by assumption.