AI crawlers

Eight crawlers decide whether your writing reaches a generative answer. Domduck reads every domain’s robots.txt on a schedule and records what it said, so the question “when did this site start blocking GPTBot” has an answer rather than a guess.

Last full reading , across 2,228 domains.

Domains read
2,228
most recent day
Catch-all block
4.0%
Disallow: / for everyone
Serve llms.txt
4.76%
the proposed standard
Agents tracked
8
three states each

How the web answers

read 2026-08-03
CrawlerBlockedAllowed by nameNot named
GPTBot
OpenAI
17.0%5.4%80.4%
OAI-SearchBot
OpenAI
6.7%5.0%91.2%
ClaudeBot
Anthropic
16.1%4.0%83.1%
PerplexityBot
Perplexity
10.6%4.8%87.4%
Google-Extended
Google
14.5%4.3%84.2%
Applebot-Extended
Apple
13.6%1.6%88.2%
CCBot
Common Crawl
17.5%2.6%83.3%
Bytespider
ByteDance
16.3%1.8%85.5%

“Not named” is a third state and not a rounding of “allowed”. A file that never mentions an agent leaves it exposed to a catch-all rule added later, so the two answers have different futures. The blocked column counts the verdict actually in force, including agents caught by a catch-all they are not named in.

What each one is for

  • GPTBot

    OpenAI’s crawler for gathering web content used to train its models.

  • OAI-SearchBot

    The crawler behind ChatGPT’s search surface: it fetches pages so they can be cited in answers.

  • ClaudeBot

    Anthropic’s crawler, used to gather web content for Claude.

  • PerplexityBot

    Perplexity’s crawler, which indexes pages so they can be cited in its answers.

  • Google-Extended

    Not a crawler at all. It is a control token: Googlebot still fetches the page, and this decides whether the content may be used for Gemini and for grounding AI Overviews.

  • Applebot-Extended

    The same shape as Google-Extended: a control token, not a crawler. Applebot fetches the page either way; this decides whether the content may train Apple Intelligence.

  • CCBot

    The crawler behind Common Crawl, a free public archive of the web that most open training corpora are built from.

  • Bytespider

    ByteDance’s crawler, gathering content used for TikTok’s search and for Doubao.

AI crawlers | Domduck