Accessibility

Baseline checks when the website changes

On the pages included in a run, axe-core checks machine-detectable WCAG issues. Checks include computable colour contrast, headings, page structure and the match between visible labels and accessible names. A keyboard pass examines reachability, focus and keyboard traps. We also check reflow at 320 CSS pixels and run an additional check at 375 pixels. The 375-pixel pass may use a documented sample of different page types. On up to four selected pages, we produce accessibility-tree transcripts and look for controls without accessible names.

In-depth review

axe-core, the keyboard pass, IBM Equal Access and Nu HTML Checker examine the pages included in a run. Siteimprove Alfa and QualWeb add rule-based checks on a sample of up to eight pages representing different page types. On that sample, we also examine:

• reflow, text resizing, increased text and line spacing, structure and language marking;
• obscured focus, reduced motion, form fields, error feedback, status messages, and open menus, dialogs and other interactive states;
• target sizes under WCAG 2.5.8, and indications that dragging, authentication, consistent help or redundant entry require further review;
• media and available captions or text alternatives; assessing their meaning and accuracy remains a separate task;
• the quality of link and control names and image descriptions: AI may support the assessment; image comparisons cover up to six content images per selected page;
• accessibility-tree reading structure and cross-browser rule checks in Chromium, WebKit and Firefox;
• link behaviour, reachable destinations and essential legal information, with additional checks for layout shifts during loading.

For up to 20 PDF links found in the included HTML pages, veraPDF checks the machine-testable requirements of PDF/UA-1. Reading order and meaningful content require additional assessment.

What a result establishes

Each run records its page coverage, sample, results and failed checks. A passed result applies to that scope; it does not establish complete WCAG conformance. Findings from different tools are compared. A finding from one tool is neither automatically disproved nor automatically confirmed. Items requiring judgement remain visible. Accessibility-tree transcripts do not replace testing with a real screen reader such as NVDA; those usability checks and human assessments are documented with their own scope. Checkout, sign-in or third-party areas that were not tested are not covered by the automated result.