AI Is Killing Wikipedia’s Human Traffic

The Wikimedia Foundation, the nonprofit that runs Wikipedia, says that shifts in how people search for information online are cutting into its human traffic.

In a blog post published today, Marshall Miller, the foundation’s senior director of product, said Wikipedia’s human visits are down about 8% over the past few months compared to the same period in 2024.

The decline was revealed after the Foundation revised how it distinguishes between human and bot traffic, something it does to better understand real readership and enforce limits on how third-party bots scrape its data for commercial search and AI tools. The update came after Wikimedia noticed what looked like a spike in human traffic from Brazil, which turned out to be mostly bots.

“We believe that these declines reflect the impact of generative AI and social media on how people seek information, especially with search engines providing answers directly to searchers, often based on Wikipedia content,” Miller wrote.

He wrote that the drop wasn’t exactly a surprise. Search engines are increasingly using AI to surface answers directly on results pages instead of linking to external sites like Wikipedia. At the same time, younger users are turning to platforms like YouTube and TikTok for information.

Unfortunately, these shifts could lead to negative ripple effects for Wikipedia. With fewer visits, Wikipedia’s volunteer base, the community that writes and edits its content, could shrink, Miller warned. And with less traffic, individual donations that keep the nonprofit running could also decline.

The situation is ironic, Miller noted, because almost all large language models (LLMS) rely on Wikipedia’s datasets for training. Yet in doing so, they may be hurting one of their most trusted sources of reliable information. Because of this, Wikimedia is urging LLMs, AI chatbots, search engines, and social platforms that use Wikipedia content to help drive more traffic back to the site.

In order to combat the issue, the nonprofit said it’s working to ensure third parties can access and reuse Wikipedia content responsibly and at scale by enforcing its policies and developing clearer attribution standards. It’s also experimenting with new ways to reach younger audiences on platforms like YouTube, TikTok, Roblox, and Instagram, via videos, games, and chatbots.

Wikimedia itself isn’t anti-AI. Just this month, the Foundation launched the Wikidata Embedding Project, a new resource that converted roughly 120 million open data points in Wikidata into a format that’s easier for large language models to use. The goal is to give AI systems access to free, higher-quality data and improve the accuracy of their answers.