The data from March through August 2025 revealed that much of the unusually high traffic in May and June came from sophisticated bots designed to evade detection, many originating from Brazil. Once these automated visitors were properly identified, the underlying trend became clear showing that real human engagement was declining. For a platform dependent on volunteer labor and individual donations, this represents more than a mere statistical concern, it signals potential structural collapse.
AI Extraction Economy
Miller’s October 2025 blog post explains how search engines are increasingly using generative AI to provide answers directly to searchers rather than linking to sites like Wikipedia. When users type questions into Google, AI Overviews now appear as boxed summaries at the top of search results that synthesize information without requiring clicks to source material.
The metrics are concerning. Data from Similarweb shows zero-click searches increased from 56% to 69% between May 2024 and May 2025. A Pew Research Center study found that just 8% of users who encountered an AI Overview clicked on a traditional search result link, compared to nearly twice as many who saw standard results. In effect, people obtain information sourced from Wikipedia without ever visiting the encyclopedia itself.
Researchers at Columbia Business School provided empirical evidence in a study published in October 2025 examining ChatGPT’s impact on Wikipedia. By comparing traffic patterns for articles whose content closely resembles ChatGPT’s output versus those that differ, they found clear evidence: readership declined noticeably for articles that mirrored ChatGPT’s output, while Wikipedia maintained advantages for complex topics where AI summaries proved inadequate.
“People are now choosing which articles to use Wikipedia for over LLMs and vice versa,” explained Hannah Li, an assistant professor at Columbia Business School who co-authored the study.
The Generational Shift
AI isn’t the only factor driving decline. Miller notes that younger generations are seeking information on social video platforms rather than the open web. According to eMarketer data, the 18-24 age segment shows a 55% generative AI internet penetration rate, with 25-34 year-olds following at 55.2%. These demographics increasingly turn to TikTok, YouTube, and Instagram for information that previously came from text-based encyclopedia searches.
This shift represents a fundamental change in knowledge consumption patterns. While Wikipedia pioneered democratizing information through collaborative text, younger users now prefer visual, algorithm-curated content optimized for engagement rather than accuracy.
The long-term trend is stark. DataReportal analysis shows organic search traffic to Wikipedia fell from approximately 5.8 billion site visits in January 2022 to just 4.3 billion in March 2025 marking a 26% decline over three years. Search engines now send 1.5 billion fewer visits to the platform than they did three years earlier.
The Existential Threat
These traffic declines threaten Wikipedia’s operational foundation. The encyclopedia operates through volunteer editors creating content while individual donors fund the nonprofit Wikimedia Foundation. Miller warns that with fewer visits to Wikipedia, fewer volunteers may grow and enrich the content, and fewer individual donors may support this work.
The risk compounds because of Wikipedia’s concentrated contributor base. Research shows that 77% of Wikipedia articles are written by just 1% of editors, creating heavy dependence on a small core group. If these editors perceive their work being consumed by AI systems without attribution, their motivation could erode. The challenge predates current concerns, the number of active contributors has been declining steadily since 2007.
Other publishers face parallel challenges. Digital Content Next found the median year-over-year decline in referred traffic from Google Search was 10% overall during May and June 2025, with losses outpacing gains two-to-one. Some sites experienced catastrophic impacts, The Planet D, a travel blog, shut down after its traffic dropped 90% following Google’s introduction of AI Overviews.
Danielle Coffey, who leads the News/Media Alliance representing more than 2,000 outlets, describes the situation plainly:
“Google is using our content without compensation, offering no meaningful way to opt out without disappearing from search entirely. It’s parasitic, it’s unsustainable, and it poses a real existential threat to many in our industry.”
The Paradox of Success
Wikipedia occupies an unusual position. Almost all large language models train on Wikipedia datasets, making the encyclopedia more influential than ever, yet increasingly invisible. When ChatGPT answers questions about historical events or scientific concepts, it draws heavily from Wikipedia’s carefully sourced articles. The encyclopedia’s knowledge reaches billions, but through intermediaries that obscure its origin.
This creates what researchers characterize as a “parasitic relationship.” AI systems extract value from Wikipedia’s volunteer-created content while contributing nothing to sustainability. Traditional search drove traffic that could convert to donations or inspire new editors. AI summaries consume Wikipedia’s intellectual commons without reciprocity.
Miller argues that companies using Wikipedia’s content “must encourage more visitors” to the website itself. The Wikimedia Foundation is developing a new framework for attributing content from the encyclopedia, working to ensure AI and search companies credit sources more visibly. Two teams, Reader Growth and Reader Experience, are testing strategies to attract audiences and improve engagement.
A Broader Warning
Wikipedia’s struggles reflect challenges throughout the digital information ecosystem. Research from SEO firm Graphite found that as of November 2024, almost half of new web articles were generated using AI, up from just 5% before ChatGPT’s launch. As synthetic content floods the web, human-curated knowledge sources become simultaneously more valuable and more vulnerable.
The encyclopedia’s situation illustrates the fragility of commons-based peer production in an AI-dominated landscape. Wikipedia succeeded through a virtuous cycle: users visited for information, some became editors, editors improved content, attracting more users. AI disrupts this by satisfying information needs without completing the loop back to the source.
As the Wikimedia Foundation works to optimize for this new reality, one thing is already clear: even the most successful collaborative knowledge project in human history faces vulnerability to disruption by technologies it helped enable. Whether Wikipedia’s decline catalyzes systemic reform in how AI systems attribute and compensate source material, or simply foreshadows gradual erosion of the open web’s knowledge infrastructure, remains uncertain.
Miller’s final message is direct:
“When you search for information online, look for citations and click through to the original source material. Talk with the people you know about the importance of trusted, human curated knowledge, and help them understand that the content underlying generative AI was created by real people who deserve their support.”
In an age of instant AI-generated answers, that may be the most radical suggestion of all.
Discover more from Being Shivam
Subscribe to get the latest posts sent to your email.