As AI eats the web, the internet’s collective memory is disappearing

That framing landed because a lot of people have already felt the failure mode in ordinary work. Journalists described using old Google tricks to find obscure scanned government PDFs that chatbots cannot surface. Developers said prior art, reference docs, and niche blog posts have become much harder to find. Several people pointed to Google’s “crawled, not indexed” behavior and the “helpful content” era as a major reason the long tail is disappearing. In that view, AI did not start the decay. It arrived after years of SEO spam, ad-driven ranking, social platforms swallowing discussion, and search engines deciding that small, specific pages were not worth indexing. The strongest throughline was that the web is not just being polluted. It is being economically hollowed out. LLMs are useful when they compress well-documented information into a fast answer, but they also break the old bargain that motivated people to publish guides, bug fixes, and niche writeups in public. If the bot gets the visit and the author gets nothing, fewer people will bother to write the next useful page. Some said they publish anyway because they want ideas to spread. More said the social reward, discoverability, and chance of collaboration were part of the point, and AI strips that out. People also pushed back on romanticizing a lost golden age. The web was already full of junk, content farms, and commercial manipulation long before LLMs. AI is best understood here as an accelerant. It raises the noise floor, makes slop cheaper, and trains users to accept plausible summaries instead of source hunting. The practical response many settled on was not to wait for Google or OpenAI to fix it. They are saving pages locally, building personal indexes, paying for alternative search, relying more on newsletters and curated sources, and expecting more knowledge to move into smaller semi-private spaces that are useful for humans but harder for bots to strip-mine.