Does blocking GPTBot in robots.txt stop ChatGPT from citing your site?

It’s easy to treat GPTBot as the switch for ChatGPT, and the same shortcut fails at the other assistants. At Anthropic, ClaudeBot is for training, while Claude-SearchBot and Claude-User serve answers. Perplexity is the exception that proves the rule: PerplexityBot is its search crawler, so blocking it does remove you from Perplexity’s answers. Across the 2,216 sites in the AI visibility study, 2.3% blocked a citation crawler, against 3.4% that blocked training only.

To check your own site, open yourdomain.com/robots.txt and sort what you find into two piles. Rules that decide whether assistants can cite you: OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot and Perplexity-User — plus any “User-agent: *” group with “Disallow: /”, which blocks all of them. ChatGPT-User and Perplexity-User may ignore the rule anyway, by their operators’ own documentation. Rules that only opt you out of training: GPTBot, ClaudeBot, CCBot, Applebot-Extended and Google-Extended — read that last one twice, because it also opts you out of Gemini grounding. The AI crawler directory covers all 25 user-agents, with the source for each.

A rule in the first pile is worth questioning: it often comes from a catch-all block or an old anti-scraper template nobody revisited. A rule in the second pile is usually a deliberate choice, and a defensible one — removing it won’t make you more citable. Being reachable by the right crawlers is necessary, not sufficient — see the other checks below.