The best trust protection is manual "vetting".

Let Sally endorse Bob as an authentic human. But only once Sally has first been endorsed trustworthy herself. Creating an opt-in circle of trust.

If a bot slips through the cracks, Sally can remove her endorsement of Bob. Which removes ALL endorsements that Bob made too, all the way down the line. Blasting bot-nets out of the sky.

Only endorse people you trust, and people you trust to make trustworthy endorsements as well.