The full list of AI crawler signatures we match.

We match 60+ answer-engine crawlers. This page is the source of truth. If your logs show one we miss, tell us — we’ll add it within 7 days and credit you.

68 signatures currently active. First-seen dates are not tracked per signature, so the table shows the date each entry was last verified against live traffic.

How we classify

Classification runs server-side on every request, before any JavaScript executes, and in three passes. The first pass looks for an exact documented user-agent token such as the ones OpenAI, Anthropic, Perplexity and Google publish. A token match is the strongest signal available and is used directly. The second pass applies a regular expression for known variants — version suffixes, vendor-specific fetcher strings, and agents that embed a product name inside a generic library string.

The third pass handles requests whose agent string is weak, generic or absent by checking the source address against published crawler IP ranges. That pass also runs as a cross-check on token matches where the vendor publishes ranges, because a request claiming a crawler identity from outside the vendor’s network is an impersonator and is reported as such. Nothing in this pipeline reads or writes anything on the visitor’s device, which is why it works without cookies or a consent banner.