WCAG 2.1 coverage: automated, assisted, manual

How much of each WCAG 2.1 Level A and AA success criterion the scan decides on its own, how much it narrows for a human reviewer, and what the EqualWeb widget repairs at run time. Generated from the rule metadata and the coverage catalog — do not edit by hand.

↩ Back to the rules index · WCAG cross-reference

Of the 50 Level A and AA success criteria in WCAG 2.1, 13 are decided entirely by the scan, 29 are assisted — the scan tests part of the criterion and shortlists the rest — and 8 need manual testing. The engine runs 156 checks in total, 27 of them live probes that drive a real browser.

Fully automated — 13 criteria#

Decided by measurement on the rendered page. The scan reports a result here, not a candidate, and the finding needs no human interpretation before it can be fixed.

Assisted — 29 criteria#

The scan machine-verifies the structural half of the criterion and shortlists the rest, with a screenshot and a code snippet per finding. What it cannot judge is meaning — whether wording, media content or visual design says the right thing.

Manual — 8 criteria#

No reliable machine signal exists. These ship as a per-page, per-round checklist in the Accessibility Audit module, with verdicts, evidence and remediation status, and they appear in the exported audit workbook.

What the EqualWeb widget repairs#

The widget works on two levels, and the scan measures both rather than assuming them.

Automatic code remediation#

Runs on every page load for every visitor, whether or not they open the widget. The repair routines are aligned one-to-one with the engine’s rules, so a finding in the report has a matching fix:

• Names and alternatives — image, button, link, SVG and object names; filename alts; AI-generated alt text · 1.1.1, 2.4.4, 4.1.2
• Forms — labels from adjacent text, dangling for targets, duplicate labels, grouped controls, autocomplete tokens, button types · 3.3.2, 1.3.5, 3.2.2
• ARIA and roles — invalid or mistyped roles, missing required states, dialog semantics, tab/menu/tree/combobox state, orphaned references · 4.1.2
• Structure — heading order, missing H1, a single main, landmarks, lists, tables, breadcrumbs · 1.3.1
• Navigation — an injected skip link, a restored focus outline, roving tabindex, keyboard-reachable scroll regions, positive-tabindex removal · 2.4.1, 2.4.7, 2.1.1, 2.4.3
• Document — page language, page title, duplicate ids, zoom-lock removal, autoplay audio, blink and marquee · 3.1.1, 2.4.2, 4.1.1, 1.4.4, 1.4.2, 2.2.2

Visitor-controlled modes#

Adaptations the end user switches on for themselves — these cover the criteria that depend on presentation:

• Contrast schemes — high contrast, monochrome, custom colours · 1.4.3, 1.4.11
• Text size and readable font · 1.4.4, 1.4.8
• Readability / text spacing · 1.4.12
• Stop animation and mute media · 2.2.2, 2.3.1, 1.4.2
• Highlight and underline links · 1.4.1
• Enlarge buttons · 2.5.8 (WCAG 2.2)
• Keyboard navigation, screen reader, virtual keyboard, reading guide — operability aids

Every page is scanned at least twice — once as delivered, once with the widget active — and each previously failing element is re-tested individually. The report shows the score with and without the widget, tags each finding as remediated or still outstanding, and separates full remediation from partial (a contrast scheme fully resolves text contrast, only partly resolves icon contrast). Findings that only a code change can resolve — keyboard traps, media alternatives, meaningful sequence — are never reported as widget-fixed.

How a criterion is classified#

A criterion is fully automated when the engine returns a definitive result on the rendered page, with no declared recall gap and no judgement about what the content means; assisted when a defined machine-verifiable subset is tested and the remainder is shortlisted for a person; and manual when no reliable machine signal exists.

Where a check documents a blind spot — only inline CSS is visible, only inline handlers are visible — its criterion is classified as assisted rather than automated, even though the check itself runs unattended. Each such limit is stated on the rule’s own page.

The Checks column counts the engine rules mapped to that criterion; the WCAG cross-reference lists them by name. site-level marks the three criteria decided by comparing scanned pages against each other rather than by a per-page rule. The counts on this page are computed from the rule metadata when the wiki is generated, and the classification is re-validated against it: adding a rule for a criterion listed as manual fails the build until it is reclassified.

Beyond Level AA the same scan also reports the WCAG 2.2 additions (2.4.11 Focus Not Obscured, 2.5.8 Target Size, 3.3.8 Accessible Authentication, 3.2.6 Consistent Help) and eight Level AAA criteria, and maps every finding to the matching clauses of EN 301 549, Section 508, IS 5568 and the ACT rules.