METHODOLOGY · UPDATED

40 of 55 WCAG 2.2 criteria, tested automatically.

Level A and AA: 26 fully, 14 partly. The other 15 need a person.

Fully tested 26Partly tested 14Needs a person 15

Fully tested

Automated checks test what the criterion asks for, on every page we scan.

Partly tested

We check part of it automatically — the note says which part. A person checks the rest.

Needs a person

No automated check today. Test it by hand, with a keyboard and a screen reader.

Every criterion, and how we test it.

  • 1.1.1 · AFully tested
    Non-text Content
    ADAGuard: images · axe-core: image-alt, image-redundant-alt
  • 1.2.1 · APartly tested
    Audio-only / Video-only
    ADAGuard: multimedia
    Native HTML + iframe detection
  • 1.2.2 · APartly tested
    Captions (Prerecorded)
    ADAGuard: multimedia
    Native + embedded player detection
  • 1.2.3 · APartly tested
    Audio Description or Alt
    ADAGuard: multimedia
    Advisory only
  • 1.2.4 · AANeeds a person
    Captions (Live)
    —
    Inherently manual
  • 1.2.5 · AAPartly tested
    Audio Description
    ADAGuard: multimedia
    Native + embedded detection
  • 1.3.1 · AFully tested
    Info and Relationships
    ADAGuard: forms, headings, tables, html structure · axe-core: label, list, th-has-data-cells
  • 1.3.2 · ANeeds a person
    Meaningful Sequence
    —
    Not checked yet (axe's tabindex rule tests 4.1.2, not reading order)
  • 1.3.3 · ANeeds a person
    Sensory Characteristics
    —
    Inherently manual
  • 1.3.4 · AAFully tested
    Orientation
    ADAGuard: mobile
  • 1.3.5 · AAFully tested
    Input Purpose
    ADAGuard: forms, mobile · axe-core: autocomplete-valid
  • 1.4.1 · AFully tested
    Use of Color
    ADAGuard: color use, links
  • 1.4.2 · AFully tested
    Audio Control
    ADAGuard: multimedia, time motion
  • 1.4.3 · AAFully tested
    Contrast (Minimum)
    ADAGuard: color contrast · axe-core: color-contrast
  • 1.4.4 · AAFully tested
    Resize Text
    ADAGuard: mobile · axe-core: meta-viewport
  • 1.4.5 · AANeeds a person
    Images of Text
    —
    Requires OCR — not automatable
  • 1.4.10 · AAFully tested
    Reflow
    ADAGuard: reflow, mobile · axe-core: meta-viewport
  • 1.4.11 · AAFully tested
    Non-text Contrast
    ADAGuard: non text contrast
  • 1.4.12 · AAFully tested
    Text Spacing
    ADAGuard: text spacing
  • 1.4.13 · AAPartly tested
    Content on Hover/Focus
    ADAGuard: hover focus content
    Confirmed F95 + JS tooltip advisory

The 15 that need a person.

Needs human judgment 12

Whether these pass depends on meaning, behaviour or several pages at once.

  • 1.2.4Captions (Live)

    Requires real-time transcription assessment — no static analysis possible.

  • 1.3.3Sensory Characteristics

    Requires understanding whether instructions rely solely on shape, color, size, location, or sound.

  • 1.4.5Images of Text

    Requires OCR to determine whether text rendered as image could have been real text.

  • 2.4.3Focus Order

    Meaningful order is a UX judgment — requires a human to navigate the actual interaction flow.

  • 2.5.1Pointer Gestures

    Requires testing real user interaction — whether multipoint/path gestures have single-pointer alternatives.

  • 2.5.2Pointer Cancellation

    Requires testing whether down-events can be aborted or reversed — behavioral, not structural.

  • 2.5.4Motion Actuation

    Requires testing whether device motion triggers can be disabled and whether alternatives exist.

  • 3.2.1On Focus

    Requires observing whether focus triggers context changes — behavioral testing only.

  • 3.2.2On Input

    Requires observing whether input triggers unexpected context changes — behavioral testing only.

  • 3.2.3Consistent Navigation

    Requires comparing navigation across multiple pages — cross-page context.

  • 3.2.4Consistent Identification

    Requires identifying components across multiple pages — cross-page context.

  • 3.3.4Error Prevention

    Requires evaluating whether legal/financial/data-loss submissions are reversible, checkable, or confirmable.

Not checked yet 3

These can be partly automated. We don’t run a check for them today.

  • 1.3.2Meaningful Sequence

    Reading order can be partly tested by comparing DOM and visual order; we do not run such a check.

  • 2.1.2No Keyboard Trap

    Needs a keyboard walk that detects focus that cannot leave a component; axe has no rule for it.

  • 2.5.3Label in Name

    axe's label-content-name-mismatch rule is experimental and off by default, so a standard scan does not run it.

A scan is not an audit. Automated checks find a large share of real problems quickly, but no tool can establish WCAG conformance on its own — the criteria that need a person still need one.

See the 40 in action on your site.