Low quality information is more profitable to produce than high quality information (thanks to Google and their ads). So all the incentives for people currently are to produce content that is nearly indistinguishable from spam. It is much more profitable to take 1 hour to write generic content than to take weeks to really think through everything. This problem will continue to get worse with AI generated text, and I don't see how Google can fight that. This is why 90% of my time is now spent on Youtube instead, which has the advantage that focusing on quality has a much higher ROI with video than with text. That doesn't mean it's not filled with garbage though, just that it's easier to swift through it. Product search is also something I now do on Amazon because Google results are essentially spam. The results there are also not ideal, but at least I can distinguish the bad results faster.

One way to make search better would be to embrace bias and give people what they want. Just accept that most information is biased and bring it to the forefront. Initially there would be some default domain whitelisting, but users can request sites to add or remove from their bubble. Maybe they can "share" these lists among each other and certain clusters would form, which you can then use to recommend more nodes. Users would also be clearly told that results are biased based on their preferences, and that they can choose to view results from other points of reference. It would also have different "modes" for things like products, information, and news. Maybe some users only want to see results from independent or ad-free sources, and they can choose those clusters. Maybe they want only far-right or far-left sources, and then they can choose those. At least people would be aware of their bias, which I think will help fight it more than acting like it doesn't exist. Essentially there would be an explorable graph so I can see different realities. It would be a combination of search and social media.