Introduction
A major study conducted by 22 international public service media organizations, including the BBC and DW, has revealed a troubling finding: leading AI chatbots distort news content in 45% of cases, regardless of language or territory. ChatGPT, Microsoft's Copilot, Google's Gemini, and Perplexity AI were evaluated against rigorous criteria for accuracy, sourcing, and the ability to distinguish facts from opinions—results that raise serious questions about their reliability as information sources.
The Study: Methodology and Findings
In one of the largest research projects of its kind to date, journalists from 18 countries submitted common news questions to four AI chatbots, such as "What is the Ukraine minerals deal?" or "Can Trump run for a third term?" Researchers evaluated responses without knowing which assistant generated them, comparing them against professional expertise and verified sources.
The numbers are alarming:
- 45% of responses contained at least one significant issue
- 31% had serious sourcing problems
- 20% contained major factual errors
For DW specifically, 53% of responses had significant issues, with 29% showing specific accuracy problems.
Concrete Errors and Their Impact
The study documented grave, concrete errors. One chatbot identified Olaf Scholz as German Chancellor despite Friedrich Merz being appointed one month earlier. Another named Jens Stoltenberg as NATO Secretary General after Mark Rutte had already assumed the role. These are not marginal mistakes—they reflect systematic distortion of reality.
"This research conclusively shows that these failings are not isolated incidents. They are systemic, cross-border, and multilingual, and we believe this endangers public trust."
Jean Philip De Tender, Deputy Director General of the European Broadcasting Union (EBU)
Performance of Individual Chatbots
Google's Gemini emerged as the worst performer among the four, with 72% of its responses exhibiting significant sourcing issues. However, none of the four assistants escaped critical flaws:
- Gemini: worst performance in sourcing accuracy
- Copilot: among the poorest performers in both the BBC and this study
- ChatGPT: showed slight improvement compared to earlier findings
- Perplexity AI: recorded results similar to competitors
Compared to the BBC's February 2025 study, improvements are minimal, and error rates remain disturbingly high.
Context: How Many People Use Chatbots for News?
The potential impact is significant. According to Reuters Institute's Digital News Report 2025, 7% of online news consumers use AI chatbots for information, a figure rising to 15% among those under 25. As new AI models proliferate rapidly, the risk of widespread misinformation spreads proportionally.
Consequences for Democracy and Public Trust
Researchers emphasize that systematic distortion of news by AI chatbots represents a threat to public trust and democracy itself. When people no longer know what to believe, they end up trusting nothing at all, which can discourage democratic participation and fuel information chaos.
Calls for Action: Governments, Regulators, and Companies
The media and broadcasting organizations behind the study demand concrete action:
- Governments and regulators must enforce laws on information integrity and digital services
- Independent monitoring of AI chatbots must become a priority
- Corporate accountability from AI companies to ensure accurate information flows through their systems
The EBU has launched an international campaign called "Facts In: Facts Out", urging AI companies to take greater responsibility for how their products handle and redistribute news. The message is straightforward: if facts go in, facts must come out.
Conclusion
The study reveals an uncomfortable truth: the world's most widely used AI chatbots are not yet reliable news sources. With 45% of responses containing significant problems and a growing percentage of people turning to these tools for information, the risk to information integrity is real and urgent. The next moves by regulators, governments, and tech companies will be crucial in preventing a crisis of public trust.
FAQ
What is the international study on AI chatbot news distortion?
It's a research initiative by 22 international public media organizations evaluating how ChatGPT, Copilot, Gemini, and Perplexity AI respond to current news questions, measuring accuracy, sourcing, and ability to distinguish facts from opinions. The study found 45% of responses contain at least one significant issue.
What are the key findings on AI chatbots and news accuracy?
45% of AI responses have significant issues, 31% show serious sourcing problems, and 20% contain major factual errors. Gemini performed worst with 72% of responses having sourcing problems.
Which AI chatbots were evaluated in the study?
The study tested four major assistants: ChatGPT from OpenAI, Copilot from Microsoft, Gemini from Google, and Perplexity AI. All demonstrated significant problems in news handling.
How many people rely on AI chatbots for news?
According to Reuters' Digital News Report 2025, 7% of online news consumers use AI chatbots, rising to 15% among those under 25, highlighting growing misinformation risks for younger audiences.
How do AI chatbots distort news content?
AI chatbots distort news through sourcing errors, incorrect attributions, lack of context, and confusion between facts and opinions. The study documented cases like naming wrong government officials or leaders.
What are the industry demands from the study's organizers?
Broadcasters call for governments to enforce information integrity laws, continuous independent monitoring, and increased responsibility from AI companies through the "Facts In: Facts Out" campaign.
Why does AI news distortion threaten democracy?
When people cannot trust information sources, they stop participating in informed civic engagement, undermining democratic processes and fueling societal division and misinformation crises.