https://openai.com/bot documents ChatGPT-User behaviour for requests generated through ChatGPT plugins. In particular:
robots.txt rules are described including:
To disallow plugins to access your site you can add the ChatGPT-User to your site’s robots.txt:
User-agent: ChatGPT-User
Disallow: /
And the egress IPs are meant to be in 23.98.142.176/28
However, when using web-requesting plugins (like BrowserOp and WebRequests) queries are being made which ignore robots.txt, use different UserAgents and which do not come from the above subnet.
BrowserOp made requests via several other IPs and User-Agents, with one request made via the OpenAI IP 23.98.142.191:
82.145.215.175 - - [05/Oct/2023:06:55:20 +0100] "GET /example/cwpathurl HTTP/1.1" 200 14637 "-" "Go-http-client/1.1"
110.238.84.125 - - [05/Oct/2023:06:55:22 +0100] "GET /example/path HTTP/1.1" 200 18319 "-" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_12_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/72.0.3626.109 Safari/537.36"
23.98.142.191 - - [05/Oct/2023:06:55:49 +0100] "GET /example/path HTTP/1.1" 200 14731 "-" "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; ChatGPT-User/1.0; +https://openai.com/bot"
WebRequests made requests via only an external IP:
76.17.123.157 - - [05/Oct/2023:07:01:14 +0100] "GET /example/path HTTP/1.1" 200 22401 "-" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/14.1.1 Safari/605.1.15"