Files
aia-website/www/reports/accessibility-audit.md
T

7.5 KiB

Accessibility Audit

Implemented

  • Semantic header, navigation, main, article, section, and footer landmarks.
  • Skip link and visible keyboard focus behavior.
  • Language and direction attributes for English, Spanish, and Arabic.
  • Native details elements for mobile navigation and FAQ accordions.
  • Form labels, autocomplete hints, disabled submission state, and explanatory text.
  • Service inquiry drawers use native modal dialogs with focus containment, Escape, backdrop, and labeled close-button dismissal plus reduced-motion support.
  • Reduced decorative image announcements through empty alt text where appropriate.
  • Responsive typography and layouts without fixed text sizing.
  • Language-switcher choices are limited to routes that actually exist.
  • The language selector uses native details/summary disclosure semantics and real links. Visitor navigation is resolved independently from SEO hreflang, so validated noindex author roots remain reachable without adding them to the hreflang graph.
  • Desktop, mobile, and footer navigation resolve against eligible generated routes. Cross-language fallbacks are explicitly identified rather than silently linking visitors to another language.
  • Arabic footer phone, email, and Latin-script address values use narrowly scoped left-to-right bidi isolation while the surrounding footer remains RTL.
  • Like buttons expose pressed and disabled state.
  • Source-faithful colors, section hierarchy, and controls were retained while avoiding the source site's modal that obscures content on initial load.

Manual Review

  • Have fluent Spanish/Arabic reviewers assess the newly localized article hero descriptions and Arabic diagram alternatives in their full article context.
  • Test keyboard and screen-reader behavior across all migrated pages.
  • Perform contrast testing against final approved brand colors.
  • Review extracted article heading hierarchy and any table markup.
  • Confirm Arabic content and RTL reading order with a fluent reviewer.
  • Review Spanish and Arabic operational labels and fallback-language notices with fluent reviewers. The source-published author labels Welcome to, All Post, and Popular Posts remain in English pending an editorial decision.
  • Test the native language disclosure, mobile navigation, search announcements, contact links, and full-page RTL reading order with screen readers. Phase 6B.5 validated generated markup and keyboard semantics, but no screen-reader executable was available in the implementation environment.
  • Validate the rebuilt mobile navigation and all full-page responsive layouts in a browser outside the autonomous sandbox.

Phase 6C.3 article-image update — 2026-10-12

The current 74-article Library corpus was re-audited after the Phase 6C.2 rewrites. All 74 featured images are informative in their present article context. The 60 records that still lacked frontmatter descriptions now have concise, locale-appropriate alt text based on direct visual inspection and the corresponding public-page evidence. No current featured image is treated as an accidental empty-alt fallback; this is a decision about the audited corpus, not a rule that future decorative images must have nonempty alt text.

The three English-text diagrams in the revised Arabic antecedent/consequence article now use separate self-hosted Arabic SVG counterparts. Each preserves the source raster and its relationships while replacing the visible labels with Arabic. The article also retains adjacent Arabic prose and semantic tables that communicate the same information without relying on the images. The causal panel order remains left-to-right where reversing it could obscure the sequence, while Arabic labels render right-to-left within their panels. At narrow widths, each detailed diagram sits in a keyboard-focusable, explicitly named horizontal scroll region with a visible focus outline. This keeps its labels at a readable minimum width without introducing page-level horizontal overflow.

Generated-markup checks cover featured-image alt propagation to the article, Open Graph, and Twitter metadata; localized Arabic asset references; and the absence of the English diagram references from the Arabic body. The SVGs were rendered locally for visual inspection at their native dimensions. A browser was not available for desktop/390-pixel overflow checks, and no screen-reader or fluent-Arabic review was performed. Those checks, plus clinical review of the translated ABA terminology and diagram relationships, remain required before cutover.

Phase 6C.4 article-link update — 2026-10-13

Generated-HTML checks now cover every Library body link's published destination, asset, and fragment, plus the rendered previous/next/related links. Spanish and Arabic self-advocacy articles identify their Halloween-guide destination as English; the two new localized sentences still require fluent-language review.

The play/leisure article retains two native image links to the full-size, self-hosted infographics. Duplicate expanded images and orphaned lightbox close symbols were removed. The ABC-data preview links to the existing full-size worksheet. These static anchors do not require JavaScript; this is markup verification, not a completed keyboard or no-JavaScript browser session.

Local Chromium failed its sandbox socket check; remote Playwright could not reach the generated site and subsequently closed its transport. Desktop/mobile visual checks, focus, image opening and browser Back behavior remain manual acceptance items. No screen-reader or fluent Arabic/Spanish acceptance is claimed. The three Phase 6C.3 Arabic diagrams still need visual-language review.

Phase 6C.5 final verification — 2026-10-13

All 74 featured images retain the individually reviewed alternatives from Phase 6C.3. Final generated-output inspection found and corrected two flattened screening/evaluation/diagnosis tables (English and Spanish), ten toilet-training H3 labels that had absorbed adjacent list prose, and the opening headings in the English/Spanish social-communication articles that had absorbed a quoted paragraph. Words, article dates and clinical guidance were not changed. The play/leisure closing CTA is now outside its last FAQ, matching the public page; all 182 FAQ answers across 30 articles match their visible content and schema.

Actual raster inspection of the three Arabic SVGs found two bounded visual issues: a replacement-behavior subtitle extended beyond its colored panel, and a few original English calm-corner title pixels appeared above the overlay. The subtitle now wraps and the calm-corner overlay starts 10 pixels higher. Reinspection shows no exposed English labels or clipped Arabic text in the inspection rasters. The previews used installed DejaVu fonts and losslessly decoded PNG backing in temporary files because this environment's librsvg did not render the SVGs' embedded WebP directly. Production SVGs retain their source WebP artwork. This is raster evidence, not browser/font-fallback acceptance.

All three diagram wrappers retain a named, focusable region, narrow-screen horizontal scrolling, a 720-pixel minimum image width and focus-outline CSS. Browser transport remains unavailable; real desktop/390px overflow, keyboard scrolling, no-JavaScript interactions, screen-reader and fluent Arabic checks remain pending. The source-inherited English tip in the Spanish positive-reinforcement article also remains a bounded language-review item. See the final Phase 6C.5 report for the route-specific checklist and release gates.