Skip to main content
Accessibility guide

WCAG 2.2 AA Checklist: What a Scanner Checks and What Needs a Human

Every WCAG 2.2 Level A and AA success criterion, grouped by principle and marked by how far an automated scanner can get: tested, assisted, or left to a person.

What this checklist is, and what it is not

WCAG 2.2 is the Web Content Accessibility Guidelines from the W3C, and Level AA is the bar most organizations work toward and most accessibility rules reference. Meeting AA means satisfying every Level A and every Level AA success criterion. This page lists all of them, grouped by the four WCAG principles, so you can work through them.

The useful part is the middle column. For each criterion it says how far an automated scanner can get: whether the machine can test it, can only assist, or cannot help at all and a person has to check. This matters because the single most common accessibility mistake is treating a clean automated scan as a pass. It is not. A scanner detects certain failures; it never confirms that a criterion is met, and a large share of WCAG cannot be judged by a machine at all.

So use this as a work plan: let a scanner clear the machine-testable failures fast, and spend your human time where the checklist says a person is required. That is where the real barriers usually hide.

How to read the middle column

Automated

A scanner can reliably detect failures of this criterion. Passing still deserves a sanity check, but the test is deterministic.

Partly automated

A scanner catches some failures or flags things to look at, but a person has to judge the rest.

Manual

A machine cannot decide this. It needs a person, often one testing with assistive technology.

These labels describe what an automated engine such as axe-core can contribute, and they are deliberately conservative. No automated tool, this one included, performs complete WCAG conformance testing.

1. Perceivable

Information and interface components must be presentable in ways users can perceive.

Success criterionLevelTestingWhat it covers
1.1.1 Non-text ContentAPartly automatedFlags missing or empty alt text; cannot judge whether the description is accurate or useful.
1.2.1 Audio-only and Video-only (Prerecorded)AManualA person confirms a transcript or alternative exists and matches the media.
1.2.2 Captions (Prerecorded)AManualA person checks captions are present, accurate and in sync.
1.2.3 Audio Description or Media AlternativeAManualHuman review of described or alternative content.
1.2.4 Captions (Live)AAManualHuman review of live captioning quality.
1.2.5 Audio Description (Prerecorded)AAManualHuman review that visual information is described.
1.3.1 Info and RelationshipsAPartly automatedDetects some structural issues (table headers, list markup, form labels); correct semantics need a person.
1.3.2 Meaningful SequenceAManualA person checks the reading order matches the meaning.
1.3.3 Sensory CharacteristicsAManualA person checks instructions do not rely on shape, size or position alone.
1.3.4 OrientationAAPartly automatedCan detect an orientation lock; edge cases need review.
1.3.5 Identify Input PurposeAAPartly automatedChecks for autocomplete attributes on common fields; correctness needs review.
1.4.1 Use of ColorAManualA person checks color is not the only way information is conveyed.
1.4.2 Audio ControlAPartly automatedCan flag auto-playing audio; a person confirms a control to stop it.
1.4.3 Contrast (Minimum)AAAutomatedText contrast against its background is measured directly. Images of text and gradients still need a person.
1.4.4 Resize TextAAPartly automatedTools assist by zooming to 200%; a person confirms nothing is lost or clipped.
1.4.5 Images of TextAAManualA person identifies text baked into images.
1.4.10 ReflowAAPartly automatedTools assist at a 320px width; a person confirms no content loss or two-dimensional scrolling.
1.4.11 Non-text ContrastAAPartly automatedFlags some UI and graphic contrast; many objects need a person to identify and judge.
1.4.12 Text SpacingAAPartly automatedTools apply the required spacing; a person checks nothing is clipped or overlapping.
1.4.13 Content on Hover or FocusAAManualA person checks hover and focus content is dismissible, hoverable and persistent.

2. Operable

User interface components and navigation must be operable.

Success criterionLevelTestingWhat it covers
2.1.1 KeyboardAManualA person operates the whole page with the keyboard alone.
2.1.2 No Keyboard TrapAPartly automatedSome traps can be detected; a person confirms focus can always leave.
2.1.4 Character Key ShortcutsAManualA person checks single-key shortcuts can be turned off or remapped.
2.2.1 Timing AdjustableAManualA person checks time limits can be extended or removed.
2.2.2 Pause, Stop, HideAPartly automatedCan flag some auto-updating or moving content; a person confirms the controls.
2.3.1 Three Flashes or Below ThresholdAManualSpecialist review of any flashing content.
2.4.1 Bypass BlocksAPartly automatedDetects a skip link or landmarks; a person confirms it actually works.
2.4.2 Page TitledAAutomatedChecks the page has a non-empty title element.
2.4.3 Focus OrderAManualA person checks focus moves in a meaningful order.
2.4.4 Link Purpose (In Context)APartly automatedFlags empty or ambiguous links; whether the purpose is clear in context needs a person.
2.4.5 Multiple WaysAAManualA person confirms more than one way to reach a page.
2.4.6 Headings and LabelsAAPartly automatedFlags empty headings and labels; whether they are descriptive needs a person.
2.4.7 Focus VisibleAAPartly automatedCan flag removed focus outlines; a person confirms a visible focus indicator.
2.4.11 Focus Not Obscured (Minimum)NEW 2.2AAManualA person checks the focused element is not hidden by sticky headers or overlays.
2.5.1 Pointer GesturesAManualA person checks multipoint or path-based gestures have a simple alternative.
2.5.2 Pointer CancellationAManualA person checks actions complete on release and can be aborted.
2.5.3 Label in NameAPartly automatedCompares the visible label with the accessible name.
2.5.4 Motion ActuationAManualA person checks motion-triggered actions have an alternative and can be disabled.
2.5.7 Dragging MovementsNEW 2.2AAManualA person checks any drag action has a single-pointer alternative.
2.5.8 Target Size (Minimum)NEW 2.2AAPartly automatedTarget dimensions can be measured; the exceptions need judgment.

3. Understandable

Information and the operation of the interface must be understandable.

Success criterionLevelTestingWhat it covers
3.1.1 Language of PageAAutomatedChecks the page declares a valid language via the html lang attribute.
3.1.2 Language of PartsAAPartly automatedDetects lang attributes on parts; whether they are correct needs a person.
3.2.1 On FocusAManualA person checks focus alone does not cause an unexpected change of context.
3.2.2 On InputAManualA person checks changing a setting does not cause an unexpected change of context.
3.2.3 Consistent NavigationAAManualA person checks navigation appears in the same order across pages.
3.2.4 Consistent IdentificationAAManualA person checks the same component is labeled the same way throughout.
3.2.6 Consistent HelpNEW 2.2AManualA person checks help is in a consistent place across pages.
3.3.1 Error IdentificationAPartly automatedCan flag some unlabeled error states; whether the message is clear needs a person.
3.3.2 Labels or InstructionsAPartly automatedFlags inputs with no label; whether instructions are sufficient needs a person.
3.3.3 Error SuggestionAAManualA person checks a helpful correction is offered when one is known.
3.3.4 Error Prevention (Legal, Financial, Data)AAManualA person checks important submissions are reversible, checked or confirmed.
3.3.7 Redundant EntryNEW 2.2AManualA person checks information already entered is not asked for again in the same process.
3.3.8 Accessible Authentication (Minimum)NEW 2.2AAManualA person checks login does not force a cognitive test with no alternative.

4. Robust

Content must be robust enough to be interpreted by a wide range of user agents, including assistive technologies.

Success criterionLevelTestingWhat it covers
4.1.2 Name, Role, ValueAPartly automatedDetects many missing names, roles and states on controls; full coverage needs a person with assistive technology.
4.1.3 Status MessagesAAPartly automatedCan flag some status patterns; correct announcement needs assistive-technology testing.

WCAG 2.2 removed the old criterion 4.1.1 Parsing. It was made obsolete because modern browsers handle the markup issues it covered, so it is not part of a 2.2 checklist.

What WCAG 2.2 added

Nine new success criteria over WCAG 2.1. Only the three at Level A or AA below affect an AA checklist; the other four are Level AAA and are listed for completeness.

CriterionLevelIn an AA checklist?
2.4.11 Focus Not Obscured (Minimum)AAYes
2.5.7 Dragging MovementsAAYes
2.5.8 Target Size (Minimum)AAYes
3.2.6 Consistent HelpAYes (Level A is required for AA)
3.3.7 Redundant EntryAYes (Level A is required for AA)
3.3.8 Accessible Authentication (Minimum)AAYes
2.4.12 Focus Not Obscured (Enhanced)AAANo, AAA only
2.4.13 Focus AppearanceAAANo, AAA only
3.3.9 Accessible Authentication (Enhanced)AAANo, AAA only

4.1.1 Parsing was removed in WCAG 2.2. Six of the nine new criteria are required for AA conformance (the three AA plus the two Level A, since AA includes all of A).

Working through it in order

A sensible order is to clear the machine-testable rows first, because they are fast, unambiguous and often the most common real failures on a page: text contrast, a missing page title, an undeclared page language, form fields with no label, missing image alternatives. An automated accessibility check handles those in one pass, and the color contrast checker handles the contrast row for a specific pair of colors.

Then spend your time on the rows marked manual, because that is where a machine is blind: keyboard operation, focus order, whether color alone carries meaning, whether an error message actually helps. Automated testing does not replace a manual accessibility review; it clears the easy failures so the review can focus on the hard ones.

Frequently asked questions

TagPipes monitors, configures and reports. It does not determine legal compliance and does not provide legal advice. Output is informational and reflects observed state at the time of observation. Rawsoft's expertise is technology, not law; please confirm any regulatory interpretation with your own counsel.

Clear the machine-testable failures in one pass.

Run a page through the free accessibility checker to catch the automated rows fast, then spend your review time where this checklist says a person is needed.

Run an automated accessibility check

Also useful: WCAG checker and the color contrast checker.