Two approaches on that.

For well-established, well-sourced, and particularly, scientific questions, authorities with a strong record of truthful and accurate accounts should both get priority and help establish truth models. Less-well-known sources which genrally agree with these also score well.

Sites showing systematic bias or poor truth valance would be discounted.

Facts, ultimately, are verified by experiment or experience. In practice, we look at reported experiential results and note which tend to agree and/or show errors which are both small and unbiased.

Those are your preferred sources.

An interesting element about lairs is that they frequently (though not always ), tend to exhibit systemic bias across multiple disciplines. Either they bullshit and make stuff up (high error, unbiased), or they show systemic biases.

Again, not always and uniformly, but frequently.

Truth and* source reputation combined give you considerable leverage.

The goal isn't perfection, but reducing the problem to tractable areas of truth valance helps greatly.

Yes, the bright line of neutrality is precious, and is way more important than correcting the tiny number of widely-accepted inaccuracies that are going to show up. The danger from a few people getting creationism results is minuscule compared to the danger that Google (further?) biases result.

The thought that google can essentially control/shape people's minds is scary.