Anyone can call themselves GPTBot. The IP address can't lie as easily.
A user agent is a line of text the client chooses, so scrapers borrow the names of well-known AI bots. Most big vendors publish the IP ranges their crawlers use. Paste an address from your logs and we'll tell you whether it belongs to one of them.
Published ranges first, reverse DNS where the vendor documents it
27 of the 169 bots in our list can be checked this way, from 10 vendors: OpenAI, Anthropic, Perplexity, Google, Microsoft, Apple, Common Crawl, Amazon, DuckDuckGo, Mistral. The vendors publish their ranges as JSON files, which we fetch and cache on our side. For Google, Microsoft, Apple and Common Crawl we can also use the reverse-DNS check each one documents: look up the hostname for the IP, check it ends in the vendor's domain, then look the hostname up again and confirm it points back to the same IP.
Meta, ByteDance and most smaller vendors publish neither, so a request claiming to be Meta-ExternalAgent or Bytespider can't be confirmed by anyone outside those companies. The crawler directory shows which bots are which.
Blocking by IP is not a reliable opt-out on its own. Anthropic, for one, says IP blocking may not work correctly or persistently and recommends robots.txt instead.
Source: Anthropic on blocking its crawlersAI crawler directory
Look up a bot from your logs
Who runs it, what it collects, the user-agent token to use in robots.txt, and how to tell a real visit from one faking its user agent.
AI visibility
Find out whether AI recommends you
Being readable is only the start. The free AI visibility check asks AI the questions your customers ask, shows the answers word for word, and tells you whether you were named.