I mean, don’t friend, or put high trust on people you don’t know is pretty strong. Due to the “six degrees of separation” phenomenon, it scales pretty easily as well. If you have stupid friends that friend bots you can cut them off all, or just lower your trust in them.

“Post-turing” is pretty strong. People who’ve spent much time interacting with LLMs can easily spot them. For whatever reason, they all seem to have similar styles of writing.

I think a web-of-trust-like network could still work pretty well where everyone keeps their own view of the network and their own view of reputation scores. I.e. don’t friend people you don’t know; unfriend people who you think are bots, or people who friend bots, or just people you don’t like. Just looked it up, and wikipedia calls these kinds of mitigation techniques “Social Trust Graphs” https://en.wikipedia.org/wiki/Sybil_attack#Social_trust_graphs . Retroshare kinda uses this model (but I think reputation is just a hard binary, and not reputation scores).

You don’t need blockchain for reputations systems, lol. Stuff like Gnutella and PGP web-of-trust have been around forever. Admittedly, the blockchain can add barriers for some attacks; mainly sybil attacks, but a friend-of-a-friend/WoT network structure can mitigate that somewhat too,