A large international study conducted by the European Broadcasting Union (EBU) in collaboration with the BBC and public broadcasters across 18 countries found that AI assistants such as ChatGPT, Microsoft Copilot and Google Gemini misrepresent news content in nearly half of their responses. The researchers evaluated over 3,000 answers across 14 languages and determined that 45% of responses had at least one significant issue—ranging from factual errors to misleading or missing sources.
Of particular concern was the high number of sourcing issues. According to the report, 72% of Gemini’s replies had major problems presenting accurate source information, including missing links or inconsistent references. The errors weren’t just superficial mistakes—they included outdated facts, fabricated details, and what the study described as “hallucinations” (cases where the AI confidently states incorrect or made-up information).
Despite the substantial error rate, the usage of AI for news is still relatively small but growing. The study cites a survey by the Reuters Institute that found around 7% of online news consumers currently use AI assistants for news, with that figure rising to about 15% among people under 25. This raises a concern because younger users appear more likely to rely on AI for news—at the same time the tools they’re using seem prone to substantial errors.
The report calls for action by all stakeholders: AI-tool developers must prioritize accuracy and transparent sourcing, news publishers should have greater control over how their content is used in AI systems, and regulators may need to consider oversight frameworks to protect news integrity and public trust.
