GPTBot

Blocking GPTBot keeps content out of future training runs and has little effect on whether ChatGPT recommends a business today. Blocking ChatGPT-User prevents the assistant from reading the site when it answers a question, so the business's pages cannot serve as current evidence.

OpenAI operates three agents with different jobs. GPTBot collects training data. ChatGPT-User fetches a specific page for a user's question. OAI-SearchBot builds the index queried by ChatGPT Search. Blocking the last two can prevent a business's own pages from appearing in live retrieval, while blocking GPTBot affects future training data.

Publishers may have editorial or commercial reasons to block training crawlers. A local service business has a different goal: it generally wants accurate descriptions of its services and location to be widely available. Decide separately whether to allow training and live retrieval because the two choices have different effects.

Request the site's robots.txt file and review its user-agent blocks. Then test the server response because robots.txt is only part of the configuration. A Cloudflare bot rule or security plugin can return a 403 response to these agents even when robots.txt allows them.

Repeat the check after CDN, firewall, or CMS security changes because those layers can alter access without changing robots.txt.

Many businesses block all three without realizing they are different, usually via a CMS toggle or a copied robots.txt. Checking takes seconds with the AI readability checker, and AI crawler covers the equivalent agents from Anthropic, Perplexity, and Google.