Why Wikipedia’s Pageviews Are Declining: The Role of AI Search and Social Video

Wikipedia has often been lauded as the last great website—an oasis on the internet dedicated to reliable knowledge rather than sensationalist content. However, even this bastion of verifiable information is now experiencing a decline. On October 18, 2025, TechCrunch reported an 8% year-over-year drop in human pageviews for the Wikimedia Foundation, attributing this trend to two significant shifts in how people seek information online: AI search summaries and social video.

The day before this announcement, the Wikimedia Foundation shared insights from its own data. Following improvements in its bot detection system this spring and subsequent reclassification of suspicious activity, the Foundation revealed a clearer picture: human visits have dropped by approximately 8% compared to the same months in 2024. This decline correlates with two primary factors: increased AI-generated answers in search results and a growing preference among younger users for social video platforms.

This issue extends beyond just Wikipedia; it touches on our search behaviors, the reliability of our information sources, and the implications of consuming knowledge in truncated formats rather than directly from the original sources. Below, we delve into the numbers, causes, and their implications for readers, publishers, and the overall health of the open web.

Understanding the Numbers

In May and June 2025, Wikipedia observed unusual spikes in apparent human traffic, particularly in Brazil. After a thorough investigation and updates to its bot detection, the Foundation discovered that a significant portion of that spike was due to sophisticated bots that had previously evaded detection. Once the data was corrected and compared year over year, human pageviews reflected a notable decline of about 8%.

Why does this decline matter? For a site like Wikipedia, visitor counts are not merely statistics. Fewer visits can lead to reduced editor engagement in improving articles, fewer newcomers learning to contribute, and a downturn in donations that support the non-profit’s operations. The Foundation has explicitly warned of these ripple effects.

The Impact of AI in Search

Over the past year, search engines have introduced prominent AI-generated summaries at the top of their results pages. These summaries are designed to answer users’ questions directly, often sourcing content from Wikipedia, YouTube, Reddit, and various news outlets. Studies are increasingly showing that when AI-generated summaries are available, users click on traditional search results less frequently.

• A July 2025 analysis from the Pew Research Center found that in the presence of Google’s AI Overview, users clicked on standard search results only 8% of the time, down from 15% on pages without AI summaries. Links within the AI summary itself were clicked only about 1% of the time.
• Ars Technica’s coverage of Pew’s findings highlighted a concerning trend: users are more inclined to end their sessions after seeing an AI summary, implying that the answer box often feels “good enough.”

These findings resonate with reports from publishers across the U.S. and Europe, who have noted that AI answer boxes decrease traffic to their sites, leading to both revenue and audience declines. Regulators are paying attention, as independent publishers have filed antitrust complaints in the EU alleging that Google’s AI Overviews unfairly divert traffic. Google has disputed these allegations.

According to Google’s own data, released on August 6, 2025, the overall volume of clicks generated from Search remains fairly stable year over year, with AI features producing slightly more “quality clicks.” The company argues that while users may not click when seeking a quick fact, they still pursue deeper information by following links, and that its AI responses aim to elevate web content through prominent citations.

Both perspectives can coexist: aggregate stability can occur alongside significant declines for specific types of sites or searches. For Wikipedia, which is central to countless fact-finding queries, the rise of immediate AI responses means many are accessing its knowledge without visiting the original source.

The Shift Toward Social Video

At the same time, the landscape of information discovery is transforming. Short-form videos are now a primary means of discovery, particularly among those under 30. According to Pew Research Center, 39% of U.S. adults under 30 regularly consume news on TikTok, a stark contrast to just 3% of those aged 65 and older. Since 2020, TikTok has seen the fastest growth as a news source among major platforms.

YouTube remains the top platform for U.S. teenagers, and many rely on social video for everything from how-tos to current events. This shift diverts attention and time away from traditional web searches. If younger users start their inquiries with content creators and short clips, they may not even open a browser tab to verify the underlying sources.

How Wikipedia is Responding

The Wikimedia Foundation is not simply raising alarms; it is taking proactive steps to adapt while maintaining the integrity of its content.

• Enforcement and Access Controls: The Foundation is developing new policies, attribution frameworks, and technical capabilities to ensure third parties access and utilize Wikipedia responsibly. Wikimedia Enterprise is part of this initiative, offering paid, reliable access to vetted data for large platforms that incorporate Wikipedia.
• User-Friendly Data for AI Developers: To mitigate harmful scraping and server overload, Wikimedia has partnered with Google’s Kaggle to provide a structured, machine-learning-friendly dataset of Wikipedia content. This initiative aims to give model builders a sanctioned, efficient alternative to scraping raw pages.
• Direct Reader Engagement: Two internal teams, Reader Growth and Reader Experience, are experimenting with how users find and interact with Wikipedia, especially on mobile devices. The Future Audiences project is exploring ways to engage younger readers on platforms they already frequent, such as YouTube, TikTok, Roblox, and Instagram.
• Human Oversight Priority: After a brief trial of opt-in AI-generated summaries on Wikipedia pages, the Foundation paused the experiment in response to significant concerns from editors regarding accuracy and trust. Community feedback emphasized apprehensions over errors, accountability issues, and the perception of putting machine-produced text above human-edited content.

The consistent message is clear: while Wikipedia is open to new ways of sharing knowledge, it relies on platforms that utilize its content to prioritize clear attribution and direct users to the original articles, supporting the volunteer ecosystem that maintains accurate and up-to-date pages.

The Stakes for the Open Web

When fewer users click through to source materials, two critical outcomes arise. First, readers risk losing essential context. AI summaries often condense, omit, or misinterpret critical nuances. If a summary is incorrect or outdated, any resulting misinformation can quickly spread. Secondly, the health of source communities is jeopardized. Wikipedia’s content is crafted by volunteers who typically begin as readers. If fewer readers choose to become contributors, the pipeline of new editors is at risk of narrowing. Additionally, since the Wikimedia Foundation relies on donations, decreased traffic could also lead to fewer contributions that support servers, tools, and community aids.

On a broader scale, the web’s value exchange is under strain. The open web thrives when discovery tools direct traffic to the content that fuels them. Many publishers report that this balance has shifted. TechCrunch has documented the effects of AI search features on publisher traffic throughout 2025, while regulators in Europe are investigating whether these new answering experiences cross competitive lines.

Evaluating Competing Claims

• Wikimedia’s Position: The Foundation reports an 8% year-over-year decline in human pageviews after adjusting for bot traffic. They attribute this to the rise of AI search answers and the allure of social video. Their warning emphasizes how fewer visits could hinder contributions and donations over time, urging platforms to prioritize direct visits and clear attribution.
• Google’s Stance: Google claims total organic clicks from Search are stable yearly and that AI Overviews create more opportunities for high-quality clicks while still providing clear external links.
• Independent Studies: Research indicates that when AI Overviews are present, users click significantly fewer links, often concluding their sessions without exploring other sites. This shift in traffic patterns could have substantial implications, even if aggregate click volumes remain unchanged.

The takeaway is straightforward: Even if general search traffic holds steady, the clicks may be diverted away from reference sources that traditionally handled quick-fact searches, placing Wikipedia at the center of this transition.

Wikipedia Knowledge Continues to Influence AI—Even Without Direct Clicks

Here lies a paradox: the less users engage with Wikipedia, the more its content shapes their experiences. Wikipedia is pivotal in training large language models and is frequently cited in AI answer boxes. Google’s own PaLM and various other models include Wikipedia in their training datasets, which means that Wikipedia remains integral to the knowledge pipeline, whether or not users visit its pages.

To harness this reality sustainably, Wikimedia is working to formalize access and attribution, utilizing Wikimedia Enterprise’s terms and the Kaggle dataset to offer structured content for researchers and startups.

Tips for Readers and Professionals: Adapting While Staying True to Facts

Whether you’re a researcher, student, marketer, or just a curious individual, this shift in information consumption habits has implications for you. Here are simple strategies to maintain accuracy and support the open web:

• Treat AI Summaries as Starting Points: Use them for orientation, but make it a habit to click through to the sources for more depth and validation. Pew data indicates how infrequently people click—make it a priority to be the exception.
• Prefer Original Content: If an AI summary references Wikipedia, open the underlying article to review citations and recent updates—it’s often just a few clicks away.
• Diversify Your Sources: For how-tos and product research, complement social video with authoritative reference pages and expert forums.
• Support Your Sources: Wikipedia’s donation-based model helps it remain independent and ad-free. If you find value in it, consider making a donation.
• For Teams Building AI Products: Consume responsibly. Use sanctioned datasets, adhere to licensing agreements, and ensure prominent attribution by linking back to the source. The Kaggle dataset exists to simplify this while reducing unnecessary scraping.

What Publishers Are Experiencing Beyond Wikipedia

Wikipedia isn’t alone in facing these challenges. Throughout 2025, various news and information sites have reported declines connected to AI functionalities in search. TechCrunch highlighted a Wall Street Journal analysis showing a broad decline in referrals coinciding with the increased prevalence of AI answer boxes. European publishers have also lodged complaints concerning both the mechanics and market impacts of AI summaries. The magnitude of this impact may differ by publisher and subject, but the trend remains: fewer clicks when AI answers dominate the top of search results.

For content that relies heavily on references, this trend is particularly pronounced. Users seeking quick definitions or dates can often get satisfactory answers from AI, resulting in a loss of visits for the sites that generated that content. Pew’s assessments of user behavior help clarify why even minor changes in click behaviors can translate to significant traffic fluctuations at scale.

The Human Element Behind Wikipedia’s Pages

The strength of Wikipedia lies in its community. Thousands of volunteers diligently write, source, and debate the information that ultimately informs your favorite search engine or chatbot. This community is vigilant about maintaining quality and skeptical of machine-generated shortcuts. In June, the Foundation briefly experimented with opt-in AI-generated summaries on selected pages, but editors raised alarms about accuracy, reliability, and the implications of prioritizing machine text over human contributions. Subsequently, the pilot was paused.

Moreover, a recent event at WikiConference North America highlighted the human aspect in a different light: volunteers worked together to neutralize an armed attendee who stormed the stage. This incident, widely reported, underscored the vulnerability and resilience of a volunteer-driven movement.

Looking Ahead

AI summaries and social video are unlikely to diminish in influence. The next phase requires establishing effective guardrails and fostering balanced value exchanges. For search engines and AI assistants, this means implementing clear linking practices, ensuring generous attribution, and creating user experiences that direct users to the sources that sustain the knowledge ecosystem. For Wikipedia, it includes continuing to adapt how people read and engage while insisting that partners treat its content with the respect it deserves.

Thus, the reported 8% decline is more than a story about Wikipedia; it serves as a warning signal for the entire landscape of the open web.

Key Takeaways

• Wikipedia’s human pageviews have dropped roughly 8% year over year, as revealed after recent data corrections.
• AI answer boxes diminish the number of clicks to external websites; when Google’s AI Overview is present, users often click fewer links and conclude sessions without visiting a source.
• Younger demographics are increasingly turning to social video as a primary news source, particularly on TikTok, impacting traditional habits of clicking through to original source pages.
• While Google asserts that overall clicks are steady and of higher quality, publishers and European regulators contest the net impact on specific segments.
• The Wikimedia Foundation is responding with initiatives that enhance accessibility for AI developers, measures to engage new readers, and a commitment to attribution while supporting its volunteer base.

FAQs

Is Wikipedia’s traffic truly down, or was it a counting error?

Both scenarios are accurate. The Foundation revealed that some of the traffic spike in May and June resulted from bots evading detection. After reclassifying that traffic, human pageviews reflect a recent year-over-year decline of about 8%.

Are AI search summaries the primary reason for Wikipedia’s click decline?

They are a significant factor but not the only one. AI answers generally reduce click-through to external sites, while social video platforms are capturing attention away from the open web—especially among younger demographics. Together, these shifts explain much of the trend Wikipedia has observed.

Does Google agree that AI search impacts website traffic negatively?

Not according to Google. They claim that overall organic clicks from Search are stable and assert that their AI features generate slightly more high-quality clicks, while emphasizing the visibility of citations in AI responses. However, independent studies and publisher feedback present a more nuanced image.

Is Wikipedia utilizing AI on its own platform?

The Foundation conducted a brief test of opt-in AI-generated article summaries in June 2025, but suspended the trial after receiving substantial community feedback. Their broader policy favors responsible AI use for support tasks rather than replacing human editing.

How can AI companies responsibly utilize Wikipedia?

Utilize sanctioned datasets such as those provided through Wikimedia Enterprise or Kaggle, adhere to licensing agreements, ensure visible credit, and facilitate user journeys that lead back to original source pages. Limiting aggressive scraping is crucial for reducing infrastructure strain while fostering responsible data use.

Conclusion

Wikipedia’s concerns are not merely nostalgic reflections on traditional web navigation. They emphasize the necessity of maintaining a feedback loop that ensures accurate, publicly accessible knowledge remains alive. AI tools and social video are now integrated into our information diets; the challenge is to transform those formats into pathways that guide users back to the source material, rather than dead ends. When platforms prioritize attribution and encourage exploration, and when readers make an effort to verify and learn, the open web can persist as a vibrant, self-correcting library rather than a static collection of summaries.

In the meantime, if an AI-generated answer has helped you today, consider visiting the page that made that answer possible. The contributions of Wikipedia’s volunteers—and the quality of the knowledge you rely on—will benefit as a result.