GPTBot

GPTBot is the user agent OpenAI uses to crawl the open web. It is one of several OpenAI agents: GPTBot is associated with training-data collection, OAI-SearchBot surfaces results in ChatGPT Search, and ChatGPT-User represents a live fetch triggered by a user action inside ChatGPT. Each has its own user-agent string and published IP ranges, and each can be allowed or blocked independently in robots.txt.

The single most important operational fact is that GPTBot fetches raw HTML and does not run JavaScript in its production paths. If your primary content is hydrated client-side, GPTBot sees an empty shell. This is the root cause of the most common AI-visibility failure we find: a beautiful React site that scores near zero on crawler readability because the answer is not in the HTML the bot receives.

Blocking GPTBot is a real strategic choice with real tradeoffs. Blocking it removes you from the data that grounds and trains OpenAI's systems, which can reduce the chance your brand is recognized and recommended later. Allowing it accepts that your content is read for training and grounding. Most brands that want AI visibility should allow the search-oriented agents (OAI-SearchBot, ChatGPT-User) at minimum. Decide deliberately and document the choice.