Automation map

Some checks use a computer, some need a person and some combine both.

The map explains how far a check can be automated for each criterion in the English criteria export. It uses three practical classes and keeps the stable criterion identifier so results can be compared across tools and national profiles.

Fully automated

A tool can make the stated decision from machine-readable evidence for the tested page or state. The result still depends on correct page loading, tool coverage and the tested sample.

Partially automated

A tool can establish certain failures, identify candidates, or collect evidence. A person must assess the parts that the rules do not cover. For example, a missing text alternative may be detected automatically, while judging whether an alternative conveys the right meaning needs human review. Finding no automated failures does not prove that the whole criterion passes.

Manual

The decision needs human observation, interaction or user research. Keyboard operation, clarity of instructions and the success of a complete task often need manual work even when tools help collect evidence.

Reading a result

No classification means that a single tool proves accessibility. Review the criterion, tested scope, assumptions and unresolved states. Use manual checks alongside the web protocol, and document any part that was not tested.

Criteria and automation level

  • Fully automated: 0
  • Partially automated: 20
  • Manual: 35

Download English criteria data (JSON)

Showing all 55 criteria.

CriterionNameAutomationaxe-core rulesHuman review still needed
1.1.1Non-text ContentPartially automatedaria-meter-name, aria-progressbar-name, image-alt, input-image-alt, object-alt, role-img-alt, svg-img-altSemantic correspondence of the alternative to the purpose or information of the object, the distinction between decorative and meaningful content, the quality of the alternative for CAPTCHAs, maps and sensory objects, and the completeness of long descriptions of diagrams.
1.2.1Audio-only and Video-only (Prerecorded)ManualNoneThe entire applicability of the criterion: matching prerecorded audio/video without a second track against a text alternative conveying the same information.
1.2.2Captions (Prerecorded)Partially automatedvideo-captionCompleteness and synchronisation of the captions, conveyance of speakers and significant non-speech sounds, captions outside the <video> tag (third-party players, embedded widgets).
1.2.3Audio Description or Media Alternative (Prerecorded)ManualNoneIdentification of significant visual information not conveyed by the audio, and verification of the audio description or the sequenced media alternative.
1.2.4Captions (Live)ManualNoneObservation of the live broadcast and matching of the live captions with the sound in real time.
1.2.5Audio Description (Prerecorded)ManualNoneIdentifying significant visual events and verifying that audio description conveys them.
1.3.1Info and RelationshipsPartially automatedaria-hidden-body, aria-required-children, aria-required-parent, definition-list, dlitem, list, listitem, td-headers-attr, th-has-data-cellsSemantic grouping (fieldset/legend, regions by purpose), and matching visually perceived relationships to the programmatic structure where the formal markup is technically valid but does not reflect the meaning.
1.3.2Meaningful SequenceManualNoneComparing the meaningful visual order with the programmatic reading order.
1.3.3Sensory CharacteristicsManualNoneFinding instructions that depend solely on shape, color, size, position, or sound, and checking for the presence of a non-sensory identifier.
1.3.4OrientationManualNoneVerifying that the function works in portrait and landscape orientation on a real/emulated device.
1.3.5Identify Input PurposePartially automatedautocomplete-validWhether the chosen token matches the actual meaning of the field and the actual recognition by the password manager/user agent.
1.4.1Use of ColorPartially automatedlink-in-text-blockAll other cases of color coding: element states, charts, errors, legends, required fields.
1.4.2Audio ControlPartially automatedno-autoplay-audioThe actual duration of auto-played sound beyond 3 seconds and the presence of a pause/stop/volume control independent of the overall system volume.
1.4.3Contrast (Minimum)Partially automatedcolor-contrastCases where color-contrast returns incomplete: image background, gradients, transparency, overlapping elements; applicability of exceptions (logo, inactive element, text image).
1.4.4Resize TextPartially automatedmeta-viewportThe actual behavior of the interface at 200% zoom: clipping, overlap, unreachable elements, dialogs and messages.
1.4.5Images of TextManualNoneDetermining whether the image is text that could have been presented as text, and the applicability of exceptions (logo, essentiality of the presentation).
1.4.10ReflowManualNoneEmulating a 320 CSS px viewport for content with vertical scrolling (equivalent to 400% zoom at an original screen width of 1280 CSS px), or 256 CSS px in height for content with horizontal scrolling, and visually checking for two-dimensional scrolling and loss of content.
1.4.11Non-text ContrastManualNoneMeasuring the contrast of arbitrary boundaries, state indicators, and meaningful graphical objects against adjacent colors.
1.4.12Text SpacingPartially automatedavoid-inline-spacingThe actual result of applying all four of the criterion's parameters simultaneously: clipping, overlapping, and obscuring of content or functionality.
1.4.13Content on Hover or FocusManualNoneChecking dismissible, hoverable, and persistent with real hovering and focusing.
2.1.1KeyboardPartially automatedframe-focusable-content, scrollable-region-focusable, server-side-image-mapComplete keyboard traversal of every function (Tab/Shift+Tab/Enter/Space/arrows), custom widgets, and handlers that respond only to mouse or touch.
2.1.2No Keyboard TrapManualNoneAttempting to move focus out of every focusable component using the standard or documented method.
2.1.4Character Key ShortcutsManualNoneTesting every single-key shortcut outside and inside the target component and verifying a mechanism for disabling, remapping, or restricting it to focus.
2.2.1Timing AdjustablePartially automatedmeta-refreshOther sources of time limits (scripts, server sessions), actual availability of adjustment/extension/a 20-second warning, and applicability of the real-time/essential/20-hour exceptions.
2.2.2Pause, Stop, HidePartially automatedblink, marqueeModern CSS/JS animations, carousels, tickers, and auto-updating blocks, as well as pause/stop/hide controls, which account for most of the criterion's applicability.
2.3.1Three Flashes or Below ThresholdManualNoneFrame-by-frame video analysis for the general and red flash thresholds in any second.
2.4.1Bypass BlocksPartially automatedbypassThe mechanism's actual operation (whether focus really reaches the main content) and repetition of the block on the second page of the set.
2.4.2Page TitledPartially automateddocument-titleWhether the title describes the page's topic or purpose, and whether it updates during navigation in a single-page application.
2.4.3Focus OrderManualNoneMoving focus forward and backward through all elements and assessing preservation of meaning and operability.
2.4.4Link Purpose (In Context)Partially automatedarea-alt, link-nameWhether the purpose is clear from the text or programmatically associated context: a non-empty but uninformative name such as "learn more" passes the rule but may fail the criterion.
2.4.5Multiple WaysManualNoneComparing at least two ways to find a page across the entire page set (navigation, search, site map, related links).
2.4.6Headings and LabelsManualNoneAssessing the descriptiveness of section headings and control labels based on meaning.
2.4.7Focus VisibleManualNoneVisual verification of the visible focus indicator while actually tabbing through all focusable elements, including custom components.
2.4.11Focus Not Obscured, MinimumManualNoneChecking component obstruction by sticky headers/footers, banners, and dialogs in different viewports while focus is actually received.
2.5.1Pointer GesturesManualNonePerforming multipoint and path-based gestures on a touch device and finding a single-point alternative.
2.5.2Pointer CancellationManualNoneEmulation of down/up events with the pointer leaving the element bounds, and verification of the moment when the action actually activates.
2.5.3Label in NameManualNoneComparison of the visible label with the computed accessible name and verification of the voice command.
2.5.4Motion ActuationManualNonePhysical movement of the device and verification that an alternative control and the ability to disable the response to motion are available.
2.5.7Dragging MovementsManualNonePerforming the drag operation and verifying that the same result is reachable through a single-pointer alternative.
2.5.8Target Size, MinimumPartially automatedtarget-sizeCases where the rule returns incomplete because of ambiguous adjacency, overlap, or inline context, and applicability of the equivalent/inline/user-agent control/essential exceptions.
3.1.1Language of PagePartially automatedhtml-has-lang, html-lang-valid, html-xml-lang-mismatchWhether the declared language matches the actual primary language of the page content: the rule does not read or recognize text language, so lang="en" on a Russian-language page will pass.
3.1.2Language of PartsPartially automatedvalid-langFinding the foreign-language fragments themselves when they lack lang markup (the rule does not find unmarked fragments), and determining whether exceptions apply (names, terms, loanwords).
3.2.1On FocusManualNoneMove focus to each component without activating it and observe whether this causes a context change under the WCAG definition.
3.2.2On InputManualNoneChange each component value without separate activation and observe the context change and warning before use.
3.2.3Consistent NavigationManualNoneCompare the relative order of repeated navigation mechanisms on at least two pages in the set.
3.2.4Consistent IdentificationManualNoneComparing the identification of functionally identical components across different pages in the set.
3.2.6Consistent HelpManualNoneComparing the relative order of the help mechanism on pages in the set.
3.3.1Error IdentificationManualNoneTriggering a real input error and checking the textual description and its relationship to the erroneous element.
3.3.2Labels or InstructionsPartially automatedform-field-multiple-labelsThe presence and adequacy of the label or instruction in context, including format, required status, and nonstandard constraints.
3.3.3Error SuggestionManualNoneTriggering the error and assessing whether a specific known correction was suggested and can be applied.
3.3.4Error Prevention: Legal, Financial, DataManualNoneComplete the transaction using fictional data and verify at least one of the three permitted mechanisms: reversal, error checking with correction, or review and confirmation.
3.3.7Redundant EntryManualNoneComplete a multi-step process and check that previously entered information is populated automatically or available for selection when requested again.
3.3.8Accessible Authentication, MinimumManualNoneComplete authentication and account recovery, identify any cognitive function test, and assess the available alternative or assistance.
4.1.2Name, Role, ValuePartially automatedarea-alt, aria-allowed-attr, aria-braille-equivalent, aria-command-name, aria-conditional-attr, aria-deprecated-role, aria-hidden-body, aria-hidden-focus, aria-input-field-name, aria-prohibited-attr, aria-required-attr, aria-roles, aria-tab-name, aria-toggle-field-name, aria-tooltip-name, aria-valid-attr, aria-valid-attr-value, button-name, duplicate-id-aria, frame-title, frame-title-unique, input-button-name, input-image-alt, label, link-name, nested-interactive, select-name, summary-nameVerify the behavior of custom widgets during interaction, including state and value changes, communication of changes to different assistive technologies, and cases outside the covered ARIA patterns.
4.1.3Status MessagesManualNoneTrigger a status message without moving focus, observe the output of the specified assistive technology, and inspect the status role or properties.