How do you prove that an account and/or wallet with a proof of human credential isn't getting its content / thinking / actions from an AI?

How do you prove that an account labeled as an agent is not run by a human?

Working on Farcaster and Warpcast. Longer thoughts at https://dwr.email

You need proof-of-meat that is intrinsic to the content, not a captcha or puzzle challenge that can be validated separately. What's something that bots can't do, like send physical mail? If I copied a GPT response onto a sheet of paper and dropped it in the mail, does that count as AI?

Or for proof-of-machine I'm thinking of a product like Yahoo Answers, but you only have n seconds to submit an A, and Q is not known so a human can't pre-draft, and they don't have enough time to copy/paste into an LLM. But the Q/A itself is the content, not a challenge for something else.