OpenAI introduced GPTBot so publishers could make an explicit choice about whether their content contributes to model training. The bot respects robots.txt rules, and OpenAI publishes its IP ranges and user-agent so sites can verify and manage access. Blocking GPTBot excludes a page from that training pipeline.
GPTBot is distinct from OAI-SearchBot and ChatGPT-User, which handle search indexing and live user-triggered browsing. Understanding these differences matters for AI visibility, because blocking one does not block the others. Acromatico maps each OpenAI agent's role so clients keep training and answer-retrieval decisions separate and intentional.
Related terms
Questions people ask
GPTBot crawls publicly accessible web pages to collect content that may help train and improve OpenAI's language models. It is the training-focused crawler, separate from the agents OpenAI uses for live search indexing or on-demand browsing when a ChatGPT user requests current information.
Add a rule to your robots.txt targeting the GPTBot user-agent with a disallow directive. GPTBot honors these instructions, and OpenAI documents its user-agent and IP ranges so you can verify requests and control access at both the robots.txt and server level.
See if AI actually names your brand
Acromatico runs live AI-visibility audits across ChatGPT, Perplexity, Gemini, and Claude.
Get a free AI Visibility Audit →