A groundbreaking international study coordinated by the European Broadcasting Union (EBU) and led by the BBC has uncovered significant accuracy issues with leading AI chatbots. The research indicates that AI assistants, increasingly used as news gateways, frequently misrepresent news content across various languages and platforms, raising serious concerns about public trust and information integrity.
Key Takeaways
- 45% of AI-generated news responses contained at least one significant issue.
- 31% exhibited serious sourcing problems, including missing or incorrect attributions.
- 20% featured major accuracy errors, such as fabricated details or outdated information.
- Google's Gemini performed worst, with 76% of its responses showing significant issues, largely due to poor sourcing.
Unprecedented Scope of Research
The intensive study involved 22 public service media organisations in 18 countries, evaluating over 3,000 responses from ChatGPT, Copilot, Gemini, and Perplexity. Professional journalists assessed the AI outputs against critical criteria like accuracy, sourcing, distinguishing opinion from fact, and providing context. The findings confirm that these failings are systemic, cross-border, and multilingual, impacting millions of users who rely on AI for information.
Growing Reliance on AI for News
AI assistants are rapidly becoming a primary source of information, with the Reuters Institute's Digital News Report 2025 showing that 7% of online news consumers use them, a figure that rises to 15% among those under 25. This reliance is concerning given the study's findings that many users trust AI summaries to be accurate, even when they are not. Errors in AI-generated news can lead to a decline in public trust in both AI developers and news providers.
Specific Performance Issues
While all tested AI assistants showed flaws, Google's Gemini was identified as the poorest performer. It struggled significantly with sourcing, providing context, and distinguishing reliable sources from satirical content. Although some improvements were noted across all platforms since earlier studies, Gemini still lags behind its peers in accuracy and reliability.
Call for Action and Future Steps
In response to the findings, the EBU and participating media organisations are urging regulators to enforce existing laws on information integrity and are advocating for ongoing independent monitoring of AI assistants. They have also released a "News Integrity in AI Assistants Toolkit" to help develop solutions and improve AI responses. A joint campaign, "Facts In: Facts Out," calls on AI companies to take greater responsibility for the accuracy and integrity of the news content they process and redistribute, demanding that "if facts go in, facts must come out."