Wikimedia says AI bots and summaries are hurting Wikipedia's traffic (www.engadget.com)

Wikimedia warned that generative AI — both LLM chatbots and AI-created summaries surfaced in search results — is reducing direct traffic to Wikipedia, threatening its role as a public, verifiable knowledge base. In a blog post, senior director Marshall Miller said updated bot-detection methods revealed an 8% year-over-year decline in page views, and Wikimedia attributes much of that drop to search engines and chatbots answering queries directly (often using Wikipedia content) and to increasingly sophisticated AI crawlers that make distinguishing human readers from bots harder. Wikimedia also noted it abandoned an internal project to auto-generate summaries after pushback from volunteer editors. For the AI/ML community this is a concrete example of downstream harms and feedback loops: models and search features that surface extracted answers can siphon attention away from source sites, reducing volunteer contributions and funding that maintain data quality. Technical takeaways include the need for better provenance, attribution, and UX that routes users back to original sources; improved bot-detection and traffic attribution practices; and consideration of long-term dataset sustainability when models rely heavily on a few public resources. Wikimedia’s call is for platforms and LLMs to explicitly surface sources and provide pathways for users to visit and participate in primary knowledge repositories.