As users increasingly use AI Agents like ChatGPT or Claude to troubleshoot, Zendesk-hosted help centers have two accessibility problems:
- Direct fetch is blocked - when an agent tries to access a help center URL, Zendesk's Cloudflare configuration returns a 403. The agent never sees the content.
- Web search returns stale content - when agents fall back to web search, they get whatever was last indexed. If a topic recently changed in our knowledge base, the agent may end up returning inaccurate, out-of-date answers from the stale search cache.
Zendesk should introduce an admin setting to host a `/llms.txt` file at the root of the help center domain. Publishing an llms.txt file would give AI agents a structured index of help center content optimized for LLM consumption.
Ideally, Zendesk would also consider allowing verified AI crawlers through Cloudflare for help center content, since this is public documentation intended to be found. At a minimum, it'd be nice if we could add instructions to the llms.txt for agents to direct them to an agent-accessible source of help content.
