Anyone tracking how AI agents actually use their selfhosted docs?

I selfhost a bunch of stuff and I’ve been wondering about something for some time that I never really cared about before. What happens when the visitor hitting the docs isn’t a person? I tried pointing a coding agent at some of my docs and giving it a basic task without any extra context. It was interesting seeing it find things I’d normally just assume or skip myself. With humans it’s pretty obvious what good docs look like. Search works navigation makes sense examples are clear, done but coding agents don’t necessarily consume docs the same way. I’m curious whether they’re finding the same pages humans do, getting stuck somewhere completely different, or just ignoring half the structure we spent time building. Has anyone here actually looked at agent traffic on their docs or logs if os what did you find and did it change anything about how you structure or host them.