AI Integration and 'Enshittification' Threaten Google Search's Reliability

Google Search is facing a crisis of reliability as the integration of artificial intelligence transforms the web from a directory of sources into a generator of answers. This shift threatens to erase the internet's role as a reliable archival record of human knowledge.

In a critical analysis published in The Walrus, author Luke Savage argues that Google Search is failing due to "enshittification" and the deployment of unreliable AI summaries. The report highlights a systemic decline in quality, noting that Google's AI summaries have begun inventing basic facts. In one confirmed instance, the system fabricated sunset times for a user in Colorado Springs, reporting that the sun had already set when it had not.

The Shift from Source to Summary

For decades, Google functioned as the primary intermediary for the web, indexing a vast ecosystem of personal websites, niche blogs, and community forums. This model allowed users to locate specific human-authored records and verify the provenance of information. However, the rise of Large Language Models (LLMs) has shifted the paradigm. Instead of directing users to a source, the engine now frequently provides a generated answer that bypasses the original creator entirely.

The Erosion of Collective Memory

This transition has significant implications for the integrity of digital information. Savage contends that when the process of searching for a source is replaced by receiving a generated summary, the provenance of information becomes obscured. This creates a risk where the "collective memory" of the internet—the ability to retrieve specific, historical, or niche human perspectives—is permanently lost.

As AI-generated content begins to flood the web, the internet risks becoming a feedback loop of summaries rather than a library of records. This environment threatens the survival of the independent, non-commercial web, as the visibility of human-authored knowledge is diminished in favor of algorithmic outputs. As Savage puts it: "When search worsens, we grumble that Google has been enshittified. When AI hallucinates, we say the model is sloppy."

The Path Forward

What remains to be seen is whether search engines can return to a model that prioritizes source transparency over generative convenience. While Google continues to integrate AI into its core product, the tension between stakeholder priorities and the need for factual accuracy remains unresolved. The industry now faces a critical question: whether the web will remain a verifiable archive of human thought or evolve into a landscape of polished but potentially hallucinated summaries.