Patterns
This part of the guide turns corpus evidence into engineering practice: concrete markup, naming, and structure patterns that serve screen reader users and AI agents from the same code.
Planned chapters, in corpus-dependency order:
- Semantic structure for two audiences
- Accessible names as machine contracts
- Forms agents can fill
- llms.txt and markdown mirrors (agent-only aids and the people-first rule that governs them)
- Anti-patterns: div soup, ARIA abuse, overlay widgets
A pattern enters this part only when it can cite a replicated corpus experiment or an established external study. Until then it lives in a chapter draft labeled as hypothesis.