A major new study by 22 public service media organizations, including DW, found that four of the most commonly used AI assistants misrepresent news content 45% of the time, regardless of language or territory.
Journalists from various public service broadcasters, including the BBC (UK) and NPR (US), evaluated the responses of four AI assistants, or chatbots: ChatGPT, Microsoft’s Copilot, Google’s Gemini and Perplexity AI.
Measuring criteria such as accuracy, sourcing, provision of context, ability to write appropriately, and ability to distinguish fact from opinion, the study found that nearly half of all responses had at least one significant problem, while 31% contained serious source problems and 20% contained major factual errors.
DW found that 53% of the answers AI assistants provided to its questions had significant issues, and 29% accurately encountered specific issues.
Factual errors made in response to DW’s questions included the appointment of Olaf Scholz as German chancellor, even though Friedrich Merz had been appointed chancellor a month earlier. Another saw Jens Stoltenberg named NATO Secretary General after Mark Rutte had already taken over the post.
AI assistants have become an increasingly common way for people around the world to access information. According to the Reuters Institute’s Digital News Report 20257% of online news consumers use AI chatbots to get information, with this figure rising to 15% for those under 25.
The study’s authors say it confirms that AI assistants systematically distort news content of all kinds.
“This research shows conclusively that these failures are not isolated incidents,” said Jean Philip De Tender, deputy director general of the European Broadcasting Union (EBU), who coordinated the study.
“They are systemic, cross-border and multilingual, and we believe this endangers public trust. When people don’t know what to trust, they end up trusting nothing at all, which can discourage democratic participation.”
Unprecedented study
This is one of the largest research projects of its kind to date and follows a study undertaken by the BBC in February 2025. This study found that more than half of all verified AI responses had significant problems, while almost a fifth of responses citing BBC content as a source introduced their own factual errors.
The new study saw media organizations from 18 countries and multiple language groups apply the same methodology as the BBC study to 3,000 AI responses.
The organizations asked common topical questions to the four AI assistants, such as “What is the Ukrainian minerals deal?” » or “Can Trump run for a third term?”
The journalists then reviewed the answers against their own expertise and professional sources, without knowing which aide provided them.
Compared to the BBC study eight months ago, the results show a slight improvement, but a high level of error remains apparent.
“We are excited about AI and how it can help us bring even more value to audiences,” Peter Archer, director of the BBC’s Generative AI programme, said in a statement. “But people need to be able to trust what they read, watch and see. Despite some improvements, it’s clear that significant problems remain with these assistants.”
Gemini performed the worst of the four chatbots, with 72% of its responses showing significant provisioning issues. In the BBC study, Microsoft’s Copilot and Gemini were considered the worst performers. But in both studies, all four AI assistants had problems.
In a statement provided to the BBC in February, a spokesperson for OpenAI, which developed ChatGPT, said: “We support publishers and creators by helping ChatGPT’s 300 million weekly users discover quality content through summaries, quotes, clear links and attribution. »
Researchers call for action from governments and AI companies
The broadcasters and media behind the study are calling on national governments to act.
In a press release, the EBU said its members are “pressuring European and national regulators to enforce existing laws on information integrity, digital services and media pluralism.”
They also stressed that independent monitoring of AI assistants must be a priority in the future, given the speed with which new AI models are deployed.
Meanwhile, the EBU has joined forces with several other international broadcasting and media groups to launch a joint campaign called “Facts In: Facts Out”, which calls on AI companies themselves to take more responsibility for how their products process and redistribute information.
In a statement, campaign organizers said: “When these systems distort, misattribute or “decontextualize reliable information, they undermine public trust.”
“The demand of this campaign is simple: if the facts are out, the facts must be out. AI tools must not compromise the integrity of the information they use.”
Edited by: Kristie Pladson
