I wonder why such discussions are always framed as an all or nothing propositions. Zero knowledge systems are a decades old invention. Just very briefly: based on some ID a site issues cryptographycally signed tokens claiming some fact, e.g. the requester being an actual real person, adulthood, etc. Such a token could be presented by an otherwise anonymous user to a 2nd site with their own signature as proof of said property in order to consume their service. Tokens could even be single use.

A requirement to prove someone is, in fact, a human is not unreasonable. Banning bots or bad actors could be a solution to a lot of the problems on social media etc...

There is naturally a major shortcoming of this scheme, authoritarians could not track people...