In this episode of Behind the Docs, Ren sits down with Dachary Carey of MongoDB to explore a rapidly emerging challenge: AI agents can’t reliably read your documentation and most teams don’t realize it.
Dachary shares how a deep dive into agent workflows (sparked by tools like Claude Code) uncovered critical gaps in how documentation is structured, delivered, and consumed by machines.
From truncated pages and hidden content to the unexpected importance of formats like llms.txt, this conversation reveals why even well-written docs can become effectively invisible to AI systems.
They also dig into:
- how agents actually interact with documentation (and where they fail)
- why technical writers are becoming essential to AI success
- and the growing gap between companies investing in docs and those cutting them entirely
If you’re thinking about AI, content strategy, or the future of technical writing, this episode is a must-listen.
Important Links:
- Dachary's LinkedIn: https://www.linkedin.com/in/dachary/
- Dachary's website: https://dacharycarey.com/
- The Agent-Friendly Documentation Spec: https://agentdocsspec.com/
- The GitHub repo for anyone who wants to contribute or leave feedback: https://github.com/agent-ecosystem/agent-docs-spec
- The npm tool (afdocs - for Agent-Friendly docs) that documentation teams can run to see how agent-friendly their documentation is: https://www.npmjs.com/package/afdocs
- The GitHub repo for that tool for anyone who finds an issue or has a suggestion or request: https://github.com/agent-ecosystem/afdocs
