Developers are increasingly discovering tools through AI interfaces like ChatGPT, Perplexity, and Claude. To get documentation cited by these systems, it needs to be structured for machine readability. Key practices include: using semantic HTML with proper heading hierarchy, writing answer-first paragraphs that are self-contained at the chunk level, adding schema.org markup (TechArticle, HowTo) with accurate dateModified fields, making code examples self-contained with comments and no magic variables, and placing an llms.txt file at the domain root to help AI agents discover key pages. Testing citability is straightforward: query AI tools with real developer questions and check which pages and sections get referenced.
Table of contents
Key TakeawaysWhat "AI Citability" Actually MeansSemantic HTML: The FoundationStructured Data Beyond the BasicsCode Blocks with ContextThe llms.txt FileAnswer-First Paragraph StructureComparing Well-Structured Documentation in the WildTesting Your Documentation's CitabilityPutting It Together485 Impressions