GPTBot (user-agent starts with `GPTBot/`) is OpenAI's primary crawler. It respects robots.txt and fetches pages for both retrieval-time grounding and model-training. A related agent, `ChatGPT-User`, fetches pages in real-time when a user asks a question that requires a fresh lookup.
Blocking GPTBot in robots.txt removes you from ChatGPT's retrieval corpus — unless you also publish an llms.txt that explicitly grants access. Allow-listing GPTBot is a cheap, high-leverage GEO action.
User-agent: GPTBot\nAllow: /\nSitemap: https://yourstore.com/sitemap.xml