Agent identity
Which requests in your log say they come from an AI provider's agent (ChatGPT-User, Claude-User, Perplexity-User, Google-Agent and others), and whether the address ranges and signing keys the provider publishes bear that out. It runs in this page: your log is never uploaded, and the page fetches nothing from the providers. Method: declared agent identity.
Spoofing cannot be checked with this site's files.
This site was built without the providers' published address ranges and signing keys, so a request that names an AI provider's agent stays a claim: nothing is verified and nothing is shown spoofed. Load copies of the providers' files under Providers' files to check against them.
Your log
A web server access log (Combined Log Format, so the user agent is there), or NDJSON with ip, user_agent and, for signatures, the request headers. Behind a proxy, log the forwarded client address, not the proxy's.
Example loaded: a synthetic log from documentation addresses, checked against synthetic range files that stand in for the providers' files. None of it is a provider's real file or a real request.
Providers' files
Without the ranges and keys a provider publishes, a user agent is only a claim: anyone can write one, and nothing can be verified or shown spoofed. This page never fetches those files itself: a request from your browser to a provider would tell it who is checking, and the site allows requests to its own files only. It starts with the copies this site was built with, if any, dated below. To check against newer files, download them from the addresses listed (or take the files swarmchase identity refresh writes into <data>/identity/) and load them here; they stay in this page.
Result
Check a log to see each record's class here.