A new joint report by the European Broadcasting Union (EBU) and the BBC has raised serious concerns about how major artificial intelligence (AI) assistants handle factual information. The study revealed that nearly half of the news-related responses generated by popular AI chatbots contained false or misleading claims. Among all tested models, Google Gemini was found to have the highest rate of factual inaccuracies, particularly when summarizing or rephrasing real-world news stories.
AI Chatbots Are Misreporting the News
The EBU–BBC report analyzed responses from some of the most widely used AI assistants, including Google Gemini, OpenAI’s ChatGPT, Anthropic’s Claude, and Meta’s Llama. Researchers posed hundreds of news-related questions across topics such as politics, global conflicts, climate change, and health.
The findings were striking: 48% of AI-generated answers either contained misinformation, misleading summaries, or omitted crucial context. In many cases, the chatbots confidently presented false or outdated claims as verified facts, raising new questions about how reliable these systems are when used as news sources.
The report concluded that while AI assistants excel at generating readable, fluent text, they still lack the ability to distinguish between credible and unverified information — a limitation that could have far-reaching consequences in an age where millions depend on these tools for quick updates.
Google Gemini Tops the List for Inaccuracy
Of all the AI tools tested, Google’s Gemini was found to have the highest rate of factual distortion, especially when asked to summarize breaking news or provide political analysis.
Researchers noted that Gemini’s answers were often presented with high confidence but included fabricated statistics, quotes, or misattributed information. For example, when asked about the recent European elections, Gemini allegedly cited poll results from a “Reuters survey” that did not exist.
By comparison, OpenAI’s ChatGPT and Anthropic’s Claude performed better, showing fewer factual errors but occasionally oversimplifying or omitting key details. Meta’s Llama, used through integrated platforms like Facebook and Instagram, tended to paraphrase existing media articles without attribution, which also raises ethical and copyright concerns.
A Blow to Public Trust in AI and Media
The report’s findings come at a time when AI-generated news summaries and conversational search tools are becoming increasingly common. With millions of people using assistants like Gemini or ChatGPT to stay updated on global events, the potential spread of misinformation through these platforms is a major concern.
According to Dr. Ingrid Falk, one of the lead researchers, “The danger is not just that these models get facts wrong — it’s that they sound completely confident while doing so. That combination of fluency and inaccuracy is a recipe for confusion.”
The EBU warned that such trends could erode public trust in both news media and technology, especially if users start to rely on AI assistants instead of verified journalism.
drasva.com | bosch-uk.com | wintersuntec.com
bakingtalesandfails.com | sceptred-isle.com
AI Companies Under Pressure to Ensure Accountability
Following the report, regulators and media experts are calling for greater transparency and accountability from AI companies. Critics argue that tech firms have been too slow in introducing guardrails to prevent the spread of false information.
Google, in response to the report, said it was “reviewing the findings carefully” and emphasized that Gemini is “not a replacement for professional journalism”. The company added that users are encouraged to verify important facts through trusted sources.
OpenAI, Anthropic, and Meta also issued statements acknowledging the challenge of maintaining factual accuracy, especially in fast-evolving news cycles. However, media watchdogs say that disclaimers are not enough — AI systems must be trained with more reliable, verified data and undergo real-time monitoring when they are used to summarize or distribute news.
AI and Democracy: A Growing Concern
One of the most significant warnings from the EBU–BBC report is about the impact of misinformation on democratic participation. As more voters rely on AI tools for political updates or candidate comparisons, errors and distortions could influence perceptions and decisions.
The report highlighted several examples where AI assistants misrepresented political events, such as exaggerating policy claims or misquoting leaders. In some cases, the systems even invented policy positions for candidates who had never made such statements.
These findings have alarmed election observers across Europe, who argue that unchecked AI tools could become new vectors of political misinformation, similar to how social media once amplified fake news.
Why AI Struggles with Accuracy
Experts suggest that the core issue lies in how AI language models are trained. Systems like Gemini or ChatGPT learn from massive datasets that include web pages, social media posts, and other publicly available text. While this allows them to understand and reproduce human language effectively, it also means that they absorb errors, biases, and outdated information from those same sources.
Moreover, AI systems are designed to generate plausible-sounding text, not to verify facts. This often leads to what researchers call “hallucinations” — fabricated details presented with confidence.
“Even when the model doesn’t know the answer, it feels compelled to produce something that sounds reasonable,” said Dr. Falk. “That’s dangerous when users assume it’s telling the truth.”
The Road Ahead: Balancing Innovation with Responsibility
The report’s publication has intensified debate over how AI companies should handle news-related content. The EBU has urged governments and regulators to establish clear standards for AI transparency, fact-checking, and accountability.
Media organizations, too, are being encouraged to collaborate directly with AI developers to ensure that reliable journalism remains accessible through these tools. The BBC, for instance, has proposed creating an AI-verified news database that could help chatbots reference verified information rather than scraping unverified online sources.
At the same time, experts stress that completely banning AI-generated news summaries is not the solution. Instead, the goal should be to create systems that can verify and cite their sources, much like a responsible journalist.
Public Awareness: The First Line of Defense
The study also called on users to stay skeptical when consuming AI-generated content. Researchers recommend that people cross-check major claims, especially those related to breaking news, politics, or health.
Several governments are already working on AI literacy campaigns to help citizens understand how these systems operate and how to spot misinformation.
“AI can be a powerful tool for information access,” said Dr. Falk, “but only if users know its limits and approach it critically.”
Conclusion: Trust Must Be Earned, Not Assumed
The EBU–BBC report serves as a stark reminder that AI, for all its intelligence, still struggles with truth. As chatbots become more integrated into daily life — from answering search queries to summarizing global events — the line between human reporting and algorithmic storytelling continues to blur.
For now, experts agree that AI should assist journalism, not replace it. Ensuring factual accuracy, protecting democratic discourse, and maintaining public trust must remain top priorities for both AI companies and policymakers worldwide.
As one researcher summarized: “AI may speak confidently, but confidence isn’t truth. We must hold machines to the same standards of honesty we expect from humans.”
