The online encyclopedia Wikimedia Foundation reports that human pageviews for Wikipedia have fallen by roughly 8 percent year-over-year, in part due to search engines and social platforms using generative-AI summaries to deliver answers directly rather than linking to the site. According to a blog post by Marshall Miller, the drop was revealed after bot-detection updates showed much of prior traffic had been non-human, and the foundation cites two major causes: search engines that now embed AI-based answers reducing outbound clicks to Wikipedia, and younger users migrating to social videos rather than open-web browsing. Meanwhile, despite being a key source for large language-models and AI tools, Wikipedia is seeing fewer direct visits—raising concerns that fewer volunteers may contribute, and donations could decline. The foundation is now working on attribution frameworks, outreach to new users, and encouraging platforms that use Wikipedia’s content to send users back.
Sources: Gizmodo, American Bazaar Online
Key Takeaways
– Because search engines and AI tools now deliver answers without sending users to the site, Wikipedia’s human-traffic has dropped about 8 percent year-over-year.
– The shift in user behavior—favoring AI summaries and social-video platforms over traditional web links—is eroding Wikipedia’s role as a gateway to information.
– Even though Wikipedia’s content powers much of the AI ecosystem, the decline in direct visits threatens the volunteer-based editing model and fundraising that support the platform.
In-Depth
It’s a quiet but significant shift in how the internet works: for over two decades, Wikipedia stood as one of the few open, trusted repositories of knowledge that people could visit, explore and link out from. But now, the very role that Wikipedia played may be changing—and not for the better, at least from its own perspective. According to recent data from the Wikimedia Foundation, Wikipedia’s human page views have dropped by roughly 8 percent compared to last year. A deeper dig revealed that earlier counts had been inflated by bots that evaded detection; once Wikipedians improved bot-filtering, a genuine decline became visible. The key culprits? Generative AI in search and a rising tide of social-video consumption.
The first major factor hinges on search engines and AI-powered services that increasingly provide answers directly on the results page, or via chatbots, instead of sending users to source sites. Wikipedia notes that many large language models and other AI systems train on or reuse Wikipedia’s content, yet those models very often deliver the answer themselves — bypassing the click to Wikipedia. The odds of someone clicking through drop when the answer is already visible. Secondly, the behaviour of younger users is shifting: less often browsing the open web via search, more often consuming short-form video and social media platforms for information. That trend further reduces traffic to traditional knowledge sites which rely on linking and browsing.
Why does this matter? For many years, the logic was circular—high traffic drove donation support, volunteer editors, and content-growth. Fewer visits mean fewer opportunities to recruit new editors or solicit micro-donations, which in turn can weaken the ecosystem of human-curated reference content. The irony is obvious: Wikipedia’s content is deeply woven into the AI fabric (training sets, LLM knowledge graphs etc), yet that same architecture now saps its visibility and perhaps its ability to sustain itself. The Foundation is responding by experimenting with a new framework of content attribution, building outreach teams aimed at younger readers, and encouraging platforms that draw upon Wikipedia’s content to drive users back for context and citation. For those of us who value a freely accessible, human-curated repository of knowledge, this moment may be one of serious consequence: if the traffic disappears, the human labour that underpins Wikipedia risks becoming harder to sustain. In short, the model that served so well may need reinvention for the AI era—unless society wants to see the free-knowledge commons quietly erode.