A recent study by the European Broadcasting Union (EBU) and the BBC revealed that leading AI assistants inaccurately present news content in almost half of their responses. The international research, which examined 3,000 responses from prominent artificial intelligence assistants in 14 languages, evaluated their accuracy, sourcing, and ability to differentiate between opinion and fact. Notably, findings showed that 45% of the AI responses analyzed contained significant issues, with 81% exhibiting some form of problem.
According to the Reuters Institute’s Digital News Report 2025, about 7% of online news consumers and 15% of individuals under 25 rely on AI assistants for news updates. In response to the study’s results, companies like Google’s AI assistant, Gemini, have expressed a commitment to enhancing their platforms based on user feedback. OpenAI and Microsoft are actively addressing issues like hallucinations, where AI models generate incorrect information due to insufficient data, while Perplexity highlights its high factuality accuracy rate on its website.
The study also identified sourcing errors in AI assistant responses, with approximately one-third displaying serious issues such as missing or incorrect attributions. Gemini, in particular, had significant sourcing problems in 72% of its responses, surpassing other assistants. Moreover, accuracy concerns, including outdated information, were found in 20% of responses across all AI assistants studied.
The study involved 22 public-service media organizations from 18 countries, including CBC and Radio-Canada, collaborating to assess AI assistants’ impact on news consumption. The EBU emphasized the importance of AI companies improving their assistants’ news-related responses to maintain public trust and democratic participation. The report called for greater accountability from AI companies, similar to the rigorous error correction processes in news organizations, to ensure accuracy and reliability in AI-generated news content.
