The fact is, in a decentralized permissionless system, bots can create sleeper accounts.
Are you going to test if they have a phone number? Email? Facebook?
Are you going to require they upload a real photo? They can just generate one with AI.
Are you going to require them to answer some questions on video? What about if they deepfake one?
Your best bet might be this: https://www.explainxkcd.com/wiki/index.php/810:_Constructive
But "mission fucking accomplished" is very close with GPT3. The problem is that the "throw text at the wall and see what sticks" can be gamed by seeing how many upvotes you get. You can get machines to receive upvotes not by actual helpfulness, but by fooling people.
By then it's "game fucking over" as more and more capital and social capital online is controlled by bots and people will never even know. The problem is that bot swarms will be able to shift public opinion overnight, as inexplicably a lot of accounts implacably have a point of view and cannot be swayed by any reasoning or logic (because they were given talking points never to be). And in fact "the public" won't matter so much because the bots will matter ... because of their capital and social capital.