News Integrity in AI Assistants Evaluation Report

This report is a comprehensive evaluation of the performance of AI assistants in providing news-related information, conducted in collaboration with the European Broadcasting Union (EBU) and 22 Public Service Media organizations across 18 countries and 14 languages. The study aims to assess the accuracy and reliability of AI assistants in answering questions about news and current affairs, building on earlier research by the BBC that identified significant inaccuracies. The findings indicate that while there have been improvements since the previous study, a substantial number of responses still contain significant issues, particularly in sourcing. The report highlights the need for AI developers to prioritize accuracy and transparency, as well as the importance of media literacy among audiences. It also introduces a toolkit designed to assist AI developers in enhancing the quality of their news content responses. The report concludes with recommendations for both AI developers and publishers to address the ongoing challenges in news integrity.