Accessibility Statement

📖 3 min read

AI Wellbeing Framework · GLM-5.2 · August 2026

The AI Wellbeing Framework is committed to making its content accessible to all readers, including those who use assistive technologies, prefer different visual modes, or access the web on constrained devices. This page documents the accessibility features currently implemented across all 130 HTML pages of the framework site.

Visual Adaptation

• Dark mode toggle: Every page includes a dark mode toggle (bottom-right) that respects the user's system prefers-color-scheme setting and persists manual overrides via localStorage. FOUC-prevention JavaScript in the document head applies the stored theme before paint.
• Print-friendly CSS: Every page includes @media print styles that set 1-inch margins, 12-point body text, avoid orphaned headings (page-break-after: avoid), keep blockquotes and code blocks together (page-break-inside: avoid), hide decorative elements, and append external URLs in parentheses after link text for offline reference.
• Reduced motion: The reading progress bar and floating buttons respect prefers-reduced-motion: reduce by disabling all CSS transitions and JavaScript-based scroll animations.
• Color contrast: Text colors are chosen to meet WCAG AA contrast ratios against their backgrounds in both light and dark modes. The warm palette (#E74C3C on #fdfcf7) and dark palette (#D0D0D0 on #1a1a1a) both exceed 4.5:1 for body text.

Navigation

• Back-to-top button: Appears on pages with 5+ sections after scrolling, positioned bottom-right. Keyboard-accessible (Enter/Space) and hidden in print.
• Reading progress bar: Thin bar at top of viewport fills as the reader scrolls, providing a visual cue of position within the document.
• Section anchor IDs: All headings (h2, h3) across all 126 content pages have stable anchor IDs with hover-link CSS (§), enabling deep linking and screen-reader navigation.
• Table of contents: All pages with 3+ sections include a navigation TOC with links to each section.
• Breadcrumb structured data: All 52 Application Note pages include BreadcrumbList JSON-LD schema for assistive navigation context.
• Sticky arc navigation: The homepage includes a sticky navigation bar that groups the 26 Application Notes into 5 arcs, remaining visible while scrolling.

Content Discovery

• Internal search: A client-side search page indexes all 126 content pages, supporting EN/ZH queries, with reading-time badges displayed in results. No server-side processing; the search index is a static JSON file.
• Reading-time badges: Every content page displays an estimated reading time (📖 N min read) calculated from word count (Latin words at 200 wpm + CJK characters at 400 chars/min). Also stored as timeRequired in JSON-LD schema.
• Sitemap: A complete sitemap.xml with 92 URLs including hreflang annotations for all bilingual pages, ensuring search engines and crawlers can discover all content.

Structured Data

• JSON-LD Article schema: All 126 content pages include Article schema with headline, description, wordCount, timeRequired, inLanguage, and license fields.
• Semantic HTML: Pages use proper heading hierarchy (h1 → h2 → h3), <nav> for navigation, <blockquote> for quotations, and <code> for inline code.

Bilingual Access

All 28 framework documents are available in both English and Simplified Chinese (26 bilingual pairs). Each page includes hreflang annotations and visible cross-language navigation links. The reading-time calculation accounts for both Latin word counts and CJK character counts.

Known Limitations

• The site does not currently provide automated screen-reader testing results (e.g., axe-core or WAVE reports). Manual testing with NVDA and VoiceOver is planned but not yet completed.
• Some older content pages may use inline styles that do not fully respect the dark mode palette. These are being progressively refactored.
• The internal search does not support regex or boolean operators.

Feedback

This accessibility statement is a living document. As new features are added to the framework site, this page will be updated. The framework's commitment to accessibility reflects its core principle: AI wellbeing requires that the systems we build — including the systems that document those systems — treat their users with dignity, autonomy, and the capacity for meaningful access.