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

521 points · 608 comments on HN · read original →

Points and comments are a snapshot, not live.

AI summaries and platform decay erode the internet's archival function.

Google's AI summaries now invent basic facts, like sunset times, signaling a breakdown of search reliability. Link rot, platform deletions, and AI pollution (e.g., content planted on Reddit to influence search answers) contaminate the public record. Disney deleted FiveThirtyEight's entire archive after laying off staff. Wikipedia traffic declines as AI ingests its content directly, reducing donations. Courts, like a German ruling holding Google liable for AI-generated false statements, and nations like France adopting privacy-preserving search Qwant, challenge Big Tech's control. The article argues for treating digital infrastructure as a public good, citing Canada's past successes (CANARIE, SchoolNet, Public Knowledge Project) as models for preserving collective memory.

The future depends on who has initiative and incentives to preserve public knowledge.

What commenters are saying

Commenters are skeptical of the article's anecdotal evidence. One notes that the user who missed a sunset because Google gave them the wrong time could have simply checked the sky. Another argues the pre-LLM Google was the best, and that relying on it for precise times was always foolish; dedicated apps exist. Several observe that while Google still yields better results than DDG for some queries (e.g., neighborhood searches), it has declined monotonically since 2008-2009, especially for finding old content and song lyrics. A minority find AI mode subjectively useful despite its inaccuracies. One commenter notes that DDG allows disabling AI features entirely, a key differentiator.