The idea is to incrementally increase the level of trust that a person is human, not verify with certainty. By having two users sign a note it's a mutual endorsement, and if someone is endorsing bot accounts they will end up destroying their own credibility when their signature shows on the bot's profile. Anyone can create accounts and sign these notes, but in a WoT model it doesn't mean anything if they don't have a following, so someone that has spent a lot of time building a following on nostr is not going to risk their credibility by verifying bots.
Discussion
But how do you know someone is endorsing bot accounts? Bots can also sign notes mutually and build up large followings.
I like the idea of the orange tick mark. Each user has to deposit some sats to show they're human, which isn't really feasible at scale for bots. Maybe some combination of this with WoT would be ideal.
It might be gameable, but I think if there were a web of endorsements and you charted it out you could gather some really interesting information and maybe infer who is trying to cheat it based on the groups that form around it.
I don't think there are any bots pretending to be human that have large followings that aren't composed mostly of bots too. Bots would endorse each other and create their own circles on the chart because they wouldn't be able to get legitimate endorsements from real honest people who are endorsing each other. Any crossover from the two groups would reveal who is a bad actor making erroneous endorsements.
That's also why clients would only show you endorsements from people you follow or have endorsed yourself. It wouldn't just be about collecting as many as you can, there has to be a real connection to be meaningful.