AI crawler access
for who.int
who.int allows every crawler that can cite it in an AI answer, while declining 1 training crawler (CCBot). That is the "cite but do not train" position.
That is below the 100% median across the 87 sites in this dataset. Resolved from https://who.int/robots.txt on 12 August 2026.
Every crawler, and the rule that decided it
A crawler named directly in its own group is governed by that group. One not named falls back to the wildcard group. Where neither exists the crawler is reported as not listed — no policy has been expressed, and in practice most crawlers will proceed.
| User-agent | Operator | Purpose | Access | Decided by |
|---|---|---|---|---|
OAI-SearchBot | OpenAI | search indexing | Not listed | no matching rule |
GPTBot | OpenAI | model training | Not listed | no matching rule |
ChatGPT-User | OpenAI | on-demand fetch | Not listed | no matching rule |
Claude-SearchBot | Anthropic | search indexing | Not listed | no matching rule |
ClaudeBot | Anthropic | model training | Not listed | no matching rule |
Claude-User | Anthropic | on-demand fetch | Not listed | no matching rule |
PerplexityBot | Perplexity | search indexing | Not listed | no matching rule |
Perplexity-User | Perplexity | on-demand fetch | Not listed | no matching rule |
Google-Extended | Gemini training control | Not listed | no matching rule | |
Bingbot | Microsoft | Copilot search index | Not listed | no matching rule |
Applebot-Extended | Apple | Apple Intelligence control | Not listed | no matching rule |
Meta-ExternalAgent | Meta | model training | Not listed | no matching rule |
Amazonbot | Amazon | model training | Not listed | no matching rule |
Bytespider | ByteDance | model training | Not listed | no matching rule |
CCBot | Common Crawl | training dataset | Blocked | direct rule |
What this does and does not tell you
It tells you what who.int states in its robots.txt, which is the file every well-behaved AI crawler consults before fetching. It is a stated preference and a voluntary protocol — not an access control, and not evidence about what any crawler actually did.
It also says nothing about whether the content is useful to an answer engine once fetched. Access is the precondition; extractable passages, visible dates, named authorship and sourced claims decide whether a reachable page is actually cited. That second layer is what the free evidence-readiness report measures.
Finally, robots.txt changes. This page reflects who.int as it was on 12 August 2026. To see the position right now, run it through the live crawler access checker.
Check your own site
The same fifteen user-agents, resolved against your live robots.txt, in about five seconds and without an account: AI crawler access checker. If the result is not what you intended, the robots.txt policy builder composes a corrected block, and how to get cited by ChatGPT explains why blocking GPTBot alone does not remove you from ChatGPT answers.
Browse the full study of 87 measured domains to see how this compares across publishers, developer references and SaaS sites.