WCAG 2.2 A and AA Suite

185 cases across 19 pages, rendered as html. 109 carry a defect and 76 are correct: flagging a correct one is a false positive.

Of those 109 defects, 33 are ones a scanner is expected to find. 12 are ones tools reasonably differ on, and 64 cannot be decided by any scanner at all: those are here to measure the tools that claim otherwise. Scoring recall against all 109 measures nothing.

Feed a scanner the list rather than crawling: https://defectlab.app/wcag22-static/urls.txt

Which case is which, per case and machine readable: https://defectlab.app/wcag22-static/cases.json

Written for DefectLab against WCAG 2.2 AA and ARIA 1.2. Not mirrored from anywhere: cases implement W3C's own catalogued Failure techniques where one exists, and say so on each case. This index is not itself a scan target and carries no defects.

Pages to scan

• images
• structure
• text
• keyboard
• pointer
• timing
• media
• links
• status
• layout
• widgets
• forms

Whole-document cases

Each of these is a page of its own: the defect is the document itself, so it cannot share a page with anything else.

• Page that reloads itself on a timer
• Document with no title element
• Document with a descriptive title
• Document with no lang attribute
• Document declaring its language
• Foreground colour specified with no background colour
• Foreground and background colours specified together