How to get your API docs cited in ChatGPT and Claude Code in 2026

Key takeaways... write the guide. Aim 1500-3000 words. Include: Key takeaways at top, H2/H3 sections, comparison table, tool embeds (Promptwatch relevant here — AI visibility, GEO, citations in ChatGPT/Claude Code — yes, directly relevant). Link Promptwatch Data reports where claims about AI search behavior: markdown report (llms.txt / markdown citations — 0.05% figure relevant to llms.txt discussion), ChatGPT citation types, Claude citation crawler visits, site: operator fanouts, average sources per response. Also link to bestgeosoftware.com or ai-rank-tools.com directory maybe once. Screenshots: the available ones are YouTube Claude API key tutorial and Apideck blog — marginal relevance. The topic is API docs cited in ChatGPT/Claude Code. The screenshots are about creating API keys — not really relevant. Skip them, or maybe skip. Rules say be selective; these are generic blog/YouTube — skip both probably. Actually the Apideck blog screenshot is a tutorial page... not authoritative for this topic. Skip screenshots.Structure:1. Key takeaways2. Why API docs are now a distribution channel (developers ask ChatGPT/Claude Code instead of Googling)3. How AI assistants actually find and cite docs (retrieval, Bing indexing for ChatGPT, Claude fewer citations, only 13% domain overlap with ChatGPT — cite Otterly study from research). Claude Code uses web search / agentic retrieval.4. Make your docs technically retrievable: allow AI crawlers in robots.txt (ClaudeBot, ChatGPTBot, GPTBot, Claude-User, etc.), llms.txt (but note markdown is 0.05% of citations — link Promptwatch markdown report), clean HTML, server-rendered, fast, no JS-only content, IndexNow, sitemaps, stable URLs.5. Write docs in citable chunks: chunk-level citability, self-contained H2 sections, direct answers first, code examples, versioned docs pitfalls.6. Claude Code specifics: agents fetch docs, MCP servers, docs optimized for agents, context efficiency.7. Offsite signals: GitHub, Stack Overflow, Reddit (note Reddit citations dropping in ChatGPT — link report), YouTube, third-party tutorials.8. Measure it: track citations, crawler logs — Promptwatch mention with embed. Comparison table of monitoring tools maybe.9. Checklist.Tools to embed: Promptwatch (definitely), maybe ZipTie (deep analysis for AI search visibility), DarkVisitors (track AI agents/bots visiting your site — very relevant!), llmtxt.info (llms.txt generator — relevant!), Otterly.AI? Keep to a handful: Promptwatch, DarkVisitors, llmtxt.info, maybe ZipTie. Also link directory bestgeosoftware.com once.Comparison table: maybe table of AI crawlers to allow in robots.txt, or table comparing ChatGPT vs Claude Code citation behavior. Do a table of AI crawlers/bots and what they do — useful. And maybe a table of monitoring tools. One table minimum; I'll do the crawler table plus a small tools table.Links to Promptwatch Data reports:- Markdown in AI Search: https://promptwatch.com/data/markdown-in-ai-search (for llms.txt section)- Reddit citations dropping: https://promptwatch.com/data/reddit-citations-are-dropping-in-chatgpt (for offsite section)- Claude citation crawler visits: https://promptwatch.com/data/claude-citation-crawler-visits-over-time (Claude retrieval)- ChatGPT site: operator fanouts: https://promptwatch.com/data/chatgpt-site-operator-fanouts (ChatGPT search behavior)- ChatGPT citation types: https://promptwatch.com/data/chatgpt-citation-types-over-time-july-2026 (docs as content type?) — could cite for what content types get cited. Documentation is a category. Fine.First Promptwatch mention: markdown link + [tool:promptwatch] embed.1001 SEO Media: not needed — topic isn't about hiring an agency. Skip.Writing style: human, opinions, varied rhythm, no banned words. Avoid em dashes mostly. No