For most flagged articles, nearly every cited sentence failed verification
Summary
Wiki Education analyzed how students used generative AI in Wikipedia editing during 2025, utilizing Pangram to detect AI-generated text. They found that over two-thirds of AI-assisted articles failed verification, meaning many sentences appeared credible but did not align with their sources. The article discusses interventions, finds AI useful for research and gap identification, and advocates broader deployment of Pangram-like tools to safeguard content integrity.