This problem is not limited to Google, it's the core value of mass-marketed LLMs, or isn't it?

Without "random comments", Google wouldn't have anything to say about "does an air purifier help my asthma, if yes: which one?" or "find the problem with this Hibernate annotation".

They also don't make much effort to exclude sloppy sites, to the contrary, they made way more efforts against SEO spam in the time when Google was a search engine, not trying to be an AI "oracle".

I think their end game is that the only metrics relevant for ranking sources are:

- agreeability (works well as a proxy for correctness with many questions!)

- originality, but not in a scientific sense, just to prevent model collapse

- legal factors such as preventing false health claims or similar things, as long as there is legislation against this kind of thing