The Walrus warns of collapsing digital memory as AI erodes search reliability
An analysis by The Walrus highlights how Google’s AI summaries are failing on basic facts while platforms like Wikipedia and the Internet Archive face existential strain, driving policy shifts in France and Germany.
An article published by The Walrus argues that Google Search is increasingly unreliable, citing instances where AI-generated summaries provided incorrect factual data, such as sunset times. The piece highlights a broader crisis in digital memory preservation, attributing the loss of collective internet history to link rot, AI hallucinations, and the removal of archives by major corporations, notably Disney’s deletion of the FiveThirtyEight archive.
The author notes that AI scraping is reducing traffic to foundational knowledge bases like the Internet Archive and Wikipedia, thereby straining their resources. Wikipedia has become the infrastructure of its own demise, with dwindling traffic meaning attention and donations no longer reliably flow back to the encyclopedia to keep it alive. Similarly, the Internet Archive is buckling under engineering strain and cyberattacks, while publishers block its crawbers out of fear that archived pages can provide AI companies with an indirect source of copyrighted material.
In response to these challenges, governments in Europe and Canada are beginning to treat information retrieval as a strategic asset. Examples include France’s adoption of the Qwant search engine and a German court ruling that held Google liable for AI-generated falsehoods. The European Parliament and French government departments have switched to Qwant, which operates on a model that does not store search data or sell personal information.
The Walrus contends that the internet's collective memory is disappearing due to these systemic failures. It highlights the strain on the Internet Archive and Wikipedia, noting that AI scraping reduces traffic to these knowledge bases. The article suggests that governments, particularly in Europe and Canada, are beginning to treat information retrieval as a strategic asset, with France adopting the Qwant search engine and a German court holding Google liable for AI-generated falsehoods.
The piece calls for public-interest digital infrastructure to preserve cultural records. It argues that we cannot aspire to sovereignty if we cannot retain and retrieve our collective memory, suggesting that public institutions should treat search as infrastructure key to digital sovereignty rather than a free consumer service. The article concludes by calling for the development of public-interest digital infrastructure to safeguard cultural records.
