How to measure AI agents traffic with Plausible?

[ItalyPaleAle]

I am well aware that this is not possible as the design of Plausible is naturally incompatible with this, but wanted to start the conversation to see what others may have found as solution/alternative. Or at least if this can be considered for future roadmap.

I have a few properties protected by Plausible, mostly docs for open source projects. I am very confident that a lot of traffic these days comes from AI agents, which make requests not just for the pages, but often fetch the content as markdown (the sites are configured to serve markdown content when negotiated).

What is the best way to measure that traffic, ideally being able to distinguish "good" AI bots from other "bad" bots (crawlers and others that add noise)? And to identify what pages they primarily land on?

I can probably find some of this data from the server's logs. However, those logs have a lot of noise, and data is separated from what Plausible reports.