Unsolvable since it's a network of fallible humans we attempt to topologically score.

You can make decent attempts, such as academic peer review. Even this system perpetuates its own problems (beta amyloid) and has perverse incentives (publish or perish, falsified results), though.

Semantic web had some good ideas about this. Networks of signed FOAF data attached to articles and posts. You could form a side graph of trust information that you could revoke at any time.

You do. I imagine people or groups curating lists of pages or sites - they decide what to put on their lists, but you decide to include them in your personal search engine or not. Or you could fork their list and edit as you see fit.