A major international study has uncovered growing concerns about AI-generated news accuracy, revealing that popular AI chatbots misrepresented factual information in 45% of responses.

The research — conducted by 22 public service media organizations worldwide — evaluated leading AI assistants including ChatGPT, Microsoft Copilot, Google Gemini, and Perplexity AI. Despite rapid progress in conversational AI, the results highlight that AI tools often fail to deliver reliable or up-to-date news, raising significant concerns about how the public consumes AI-curated information.

AI-Generated News Accuracy: The Numbers Behind the Study

Researchers examined AI-generated responses based on four essential criteria — accuracy, sourcing, context, and the ability to separate fact from opinion.

Here’s what they found:

• 45% of responses contained at least one significant inaccuracy.
• 31% showed serious sourcing problems.
• 20% had factual errors or outdated data.

A striking example involved an AI chatbot identifying Olaf Scholz as Germany’s Chancellor, even though Friedrich Merz had already taken office. Another case falsely attributed NATO’s leadership to Jens Stoltenberg instead of Mark Rutte.

These examples highlight why AI-generated news accuracy remains a serious issue, especially when users rely on chatbots for real-time global updates.

Public Trust at Risk: Can AI Be Trusted for News?

According to the Reuters Institute Digital News Report 2025, about 7% of online users now rely on AI chatbots for news — a figure that jumps to 15% among people under 25.

Experts warn this trend could erode public trust in journalism. Jean Philip De Tender, Deputy Director General of the European Broadcasting Union (EBU), cautioned:

“Uncertainty in trust leads to a complete lack of trust — and that’s dangerous for democracy.”

As AI becomes more integrated into media platforms, ensuring AI-generated news accuracy will be vital for maintaining both credibility and civic engagement.

Inside the Study: How Researchers Tested the Chatbots

The large-scale study followed the same method as a BBC investigation conducted in February 2025. Journalists asked real-world news questions to the four major AI assistants, without knowing which chatbot was providing the answers.

The outcome revealed minimal improvement in AI-generated news accuracy compared to previous studies.

Notably:

• Google Gemini had the highest rate of sourcing errors — 72% of its responses were problematic.
• Microsoft Copilot also struggled with factual consistency.
• ChatGPT and Perplexity AI performed slightly better but still produced noticeable factual mistakes.

The findings indicate that AI chatbots continue to face major challenges in maintaining factual integrity across languages and topics.

Media Industry Demands AI Accountability

Following the report, the EBU and multiple broadcasters urged stronger regulation to ensure AI systems uphold factual standards when handling journalistic content.

The EBU also launched a global campaign called “Facts In: Facts Out”, emphasizing that AI companies must guarantee accuracy and transparency when processing verified news.

Their statement reads:

“If facts go in, facts must come out.”

This marks a growing movement demanding AI-generated news accuracy and ethical responsibility in how large AI systems process and distribute information.