Artificial intelligence chat assistants distort or misreport news content in almost half of their answers, according to new research released Wednesday by the European Broadcasting Union (EBU) and the BBC.
AI Assistants Found to in Nearly Half of Responses, EBU–BBC Study Reveals
The study analyzed 3,000 responses from top AI-powered assistants — including Open AI's ChatGPT, Microsoft’s Copilot, Google’s Gemini, and Perplexity — to assess factual accuracy, source attribution, and the distinction between fact and opinion.
Covering 14 languages, the research found widespread inconsistencies, raising concerns about the reliability of AI tools as news sources. Media experts and regulators have increasingly warned about the potential spread of misinformation through generative AI systems.
According to the findings, 45% of AI-generated responses contained at least one major error, while 81% had some form of problem. The EBU and BBC emphasized the urgent need for greater transparency in how AI systems handle and present news, warning that their growing use could blur the boundaries between verified journalism and synthetic content.
A third of responses showed significant sourcing issues — such as missing, misleading, or incorrect citations. Google’s Gemini had the highest rate of sourcing problems at 72%, compared to under 25% for the other assistants.
Accuracy issues were detected in around 20% of all responses, including outdated or false information. Examples cited included Gemini’s incorrect statement about changes to disposable vape laws and ChatGPT referring to Pope Francis as alive months after his reported death.
The research involved 22public-service media organizations across 18 countries, including France, Germany, Spain, Ukraine, the U.K., and the U.S.
As more people turn to AI assistants instead of traditional search engines for news, experts warn public trust may erode.
“When people don’t know what to trust, they end up trusting nothing — and that can discourage democratic participation,” said EBU Media Director Jean Philip De Tender.
According to the Reuters Institute’s Digital News Report 2025, 7% of online news consumers — and 15% of those under 25 — now rely on AI assistants for news updates.
The report urged AI developers to take responsibility and improve how their systems deliver and cite news information.
No comments:
Post a Comment