There’s a reason that works (sorta, mostly, and when it does).
The infrastructure of web navigation and search is foundationally built on textual, first party published content that enjoys some presumptions of good faith authorship: that it was written to be read, by a reader, and meant to be informative, persuasive/argumentative, or recreational/entertaining.
Think: journal articles, Usenet debates, faq compilations, RFCs, product spec sheets, forum postings, IRC logs, ur-blogs, even the original social media content types (tweets, wall posts, comments).
The cycles and epicycles of architecting atop this “real” content gives us mechanisms that reward and incentivize content produced neither for the ends of the writer writing nor any reader reading, but rather in service of manipulating search and driving ux patterns of various darkness.
All the good stuff on the internet is reborn in a spasm of returning to people creating content they care about for other people who care. Then eventually with success it gets polluted into shit with bots, seo, mass media and agencies, influencers, etc.
One silly rubric I use is: could this page / app / venue have been there in 1999? If so it’s probably worth reading. If not, who knows.
(Of course there are cesspools that qualify too, like the chan world and certain subreddits, but there were cesspools in the early days of the web as well.)