AI and Social Media Dwindle Wikipedia’s Human Traffic, Warns Wikimedia Foundation

The Wikimedia Foundation, which operates Wikipedia, has reported a significant decline in human traffic to the site, attributing the drop to the growing influence of generative AI and shifting online behaviors. In a blog post, Marshall Miller, the foundation’s senior director of product, revealed that human visits to Wikipedia have fallen by approximately 8% over recent months compared to the same period in 2024. This decline became apparent after Wikimedia updated its methodology for distinguishing between human users and automated bots—critical for accurately measuring real reader engagement and enforcing rules against unauthorized scraping of content by third-party AI and search tools. The change uncovered a previous misclassification: a spike in traffic from Brazil, once thought to reflect increased human interest, was largely driven by bots. Miller explained that the drop in traffic aligns with broader trends in how people access information. Search engines are increasingly using AI to generate direct answers on search result pages, often pulling from Wikipedia content without linking back to the original source. This reduces the need for users to click through to Wikipedia, directly impacting its visibility and readership. At the same time, younger audiences are turning to platforms like YouTube and TikTok for quick, visually engaging information, further diverting traffic away from traditional encyclopedic sources. The foundation warns that this trend could have serious consequences. Wikipedia relies heavily on a global community of volunteer editors to maintain and improve its content. Fewer visitors may lead to reduced engagement from these volunteers, potentially weakening the quality and sustainability of the platform. Additionally, lower traffic could translate into fewer individual donations, which are vital to funding the nonprofit’s operations. The situation presents a paradox: while Wikipedia remains one of the most important training sources for large language models (LLMs), the very AI systems built on its data may be contributing to its decline. To address this, Wikimedia is calling on AI developers, search engines, and social media platforms to help drive more traffic back to Wikipedia by properly attributing content and linking to original sources. To support responsible use of its data, the foundation is strengthening enforcement of its reuse policies and developing clearer standards for attribution. It is also exploring new ways to connect with younger audiences through platforms like YouTube, TikTok, Roblox, and Instagram—using videos, interactive games, and chatbots to make knowledge more accessible and engaging. Importantly, Wikimedia is not opposed to AI. In fact, this month it launched the Wikidata Embedding Project, which transforms about 120 million open data points from Wikidata into a machine-readable format optimized for LLMs. The goal is to provide AI systems with high-quality, free data while encouraging them to cite and direct users back to Wikipedia, fostering a more sustainable relationship between the platform and the AI ecosystem.