Wikipedians Believed AI Would Come and Go. They Were Wrong
Richard Cooke | The Big Story | October 13, 2026
In a world of AI Overviews, Wikipedia usage declines. But the real loss may be far more existential than website traffic.
When the Experts Underestimated the Machine
For much of the past decade, the Wikipedia community treated artificial intelligence as a passing fad. Editors and administrators—people deeply invested in the project's long-term survival—assumed that large language models would hit a ceiling, plateau in usefulness, and eventually fade from public attention. That assumption now looks less like cautious skepticism and more like a miscalculation with consequences that reach far beyond page views.
By 2026, AI systems have not come and gone. They have reshaped how hundreds of millions of people find, consume, and trust information. And Wikipedia—the largest collaboratively built knowledge base in human history—sits squarely at the pivot point of that transformation.
The Traffic Mirage
The most visible symptom is the decline in Wikipedia usage. Search engines increasingly surface AI-generated summaries instead of linking users directly to source pages. AI Overviews, conversational assistants, and chat-based interfaces deliver condensed answers without requiring a click-through. For a site built on the idea that readers will arrive, browse, and perhaps contribute, this shift is structural, not cyclical.
But traffic metrics may be the least important thing at stake. Wikipedians have weathered slumps before. What is different now is the erosion of the feedback loop that sustains the project: readers become editors, editors become administrators, and the whole system renews itself through visible participation.
More Than a Website
If Wikipedia fades from everyday view, what disappears is not just a popular reference site. It is a working model of decentralized, transparent, and accountable knowledge production—one that AI systems have quietly relied on for years.
Large language models are trained on vast corpora, and Wikipedia has long been one of the highest-quality, most consistently structured sources available in the open web. When users ask an AI assistant a question, the answer may be traceable back to a Wikipedia article—sometimes verbatim, often in paraphrase, and almost never with clear attribution.
This creates a paradox: the more AI mediates access to knowledge, the more the underlying sources matter, and the less visible they become. Wikipedia's influence expands even as its direct audience contracts.
What Is Actually Lost
The existential risk is not that Wikipedia disappears overnight. It is that the incentives and rewards that have kept it healthy begin to decay. Volunteer editors need to see that their work is read, that it matters, and that it contributes to something shared. When AI systems absorb that work without returning traffic, credit, or contribution, the generous spirit that built the encyclopedia is put under strain.
At the same time, AI-generated misinformation and fabricated citations have made the verification role of Wikipedia more important than ever. In a web increasingly saturated with synthetic content, a source that is transparent about its edits, sourced in public, and governed by community norms becomes not obsolete but essential.
A Fragile, Irreplaceable Commons
The Wikipedians who believed AI would come and go were not naive. They were applying a reasonable historical pattern: technologies rise, peak, and recede. What they missed was the degree to which AI would become infrastructure rather than novelty—a layer beneath the web rather than a competitor to it.
That shift demands a new response. Attribution, licensing, and fair compensation for training data are no longer academic debates. They are practical questions about whether the open knowledge commons can survive in a world where machine intermediaries capture the value of human-curated information.
Wikipedia's future may depend less on whether people visit the site and more on whether the broader ecosystem recognizes what is being lost when they stop.
via Wired AI