A recent study by the European Broadcasting Union (EBU) and the BBC revealed that leading AI assistants inaccurately represented news content in almost half of their responses. The international research, which analyzed 3,000 responses from AI assistants in 14 languages, including OpenAI’s ChatGPT, Microsoft’s Copilot, Google’s Gemini, and Perplexity, found that 45% of the responses contained significant issues, with 81% exhibiting some form of problem.
According to the Reuters Institute’s Digital News Report 2025, approximately 7% of online news consumers and 15% of individuals under 25 rely on AI assistants for news consumption. The companies behind these AI assistants, including Gemini, Google, OpenAI, and Microsoft, have acknowledged the need for improvement in addressing inaccuracies and misinformation.
One of the key findings of the study was that a third of AI assistant responses contained serious sourcing errors, such as missing or incorrect attribution. Gemini, in particular, showed significant sourcing issues in 72% of its responses. Additionally, 20% of responses from AI assistants were found to have accuracy issues, including outdated information.
The study cited examples of misinformation, such as Gemini providing incorrect information about changes to a law on disposable vapes and ChatGPT erroneously stating that Pope Francis was still alive months after his passing. The EBU emphasized the importance of AI companies enhancing the accuracy and accountability of their assistants when responding to news queries to maintain public trust in the information provided.
In response to the study, the EBU called for AI companies to adopt more stringent measures to ensure the reliability of news-related responses from their assistants, similar to the rigorous error correction processes followed by news organizations. The EBU emphasized the significance of accountability in maintaining trust and fostering democratic engagement among users of AI assistants.
