AI chatbots like ChatGPT and Gemini may be contributing to misinformation

AI news takes a hit on accuracy

A major EBU and BBC study found that 45% of AI-generated news responses contained significant issues such as factual mistakes, sourcing problems, or confusion between fact and opinion.

Researchers tested ChatGPT, Copilot, Gemini, and Perplexity on current news stories, revealing that nearly half of their answers were flawed.

These were not minor errors but deep problems in accuracy and verification. The results raise questions about how much users can trust AI tools to deliver reliable information in fast-changing news environments.

Inside the global accuracy test

Researchers from the EBU and BBC tested over 3,000 AI-generated answers from four major assistants across several languages. Each was asked identical news questions and compared against verified sources. Nearly half of the results contained major issues.

Common errors included outdated statistics, incorrect names, and missing context. The findings showed that while AI tools can summarize quickly, they often misrepresent or misinterpret complex events. The study proved that confident presentation does not equal reliable information when it comes to AI-generated news.

BBC finds factual flaws in AI-generated news

People expect AI assistants to deliver fast and accurate news, but reality falls short. The EBU and BBC found that 20 percent of responses contained clear factual errors, such as wrong dates, names, or timelines, and 31 percent had serious sourcing problems, including missing or incorrect citations.

These results show that even the most advanced chatbots struggle with understanding context. The issue isn’t about speed or grammar but truth. When AIs distort details instead of clarifying them, they undermine confidence in the technology meant to keep people informed.

The common types of AI mistakes

The EBU study found that AI news tools made a range of errors, from simple factual mistakes to complete fabrications. Some gave outdated names or dates, while others invented quotes or cited fake news outlets.

One BBC-cited example showed an AI claiming to quote a person who had never made that statement. These issues are not isolated; they appear across all major chatbots. The pattern suggests a structural problem in how large language models process, infer, and rewrite factual information.

When bad AI news causes real harm

Misinformation generated by AI assistants has broader consequences than just individual mistakes. The EBU and BBC warned that inaccurate news erodes public trust and may discourage participation in democratic debate.

While some reports suggest that a single factual error can sharply reduce trust in AI, the broader issue is credibility. When people cannot separate truth from error, confidence in both media and technology suffers. The study concluded that factual accuracy is essential to maintain trust in AI systems.

Why AI assistants sound so certain?

AI assistants are built to sound confident even when they are wrong. Unlike a search engine that offers multiple sources, they generate one polished answer. Developers admit that current systems are rewarded for providing a complete-sounding response rather than saying they do not know.

This leads to overconfident delivery of inaccurate information. The EBU’s toolkit for improving AI reliability now encourages developers to train models that admit uncertainty. Confidence, as the study showed, is not a substitute for truth.

Big names, big accuracy gaps

The EBU and BBC study found clear differences in performance among popular AI systems. Google’s Gemini performed worst, with 76% of its news answers containing significant issues, mostly due to unreliable sources.

ChatGPT and Microsoft’s Copilot also showed weaknesses but performed better than Gemini overall. Gemini’s sourcing errors reached 72%, the highest rate in the study.

These findings highlight that even the largest, best-funded AI tools struggle to maintain factual consistency in real-world information tasks.

When facts blur into opinions

AI assistants often confuse opinion with fact, a flaw identified in the EBU and BBC analysis. In one example, a chatbot presented an established legal verdict as a personal viewpoint. Such cases show how easily AIs merge subjective tone with factual reporting.

Researchers warned that this blurring changes how readers perceive truth, since it replaces clear reporting with emotionally colored summaries. These subtle distortions are among the most dangerous because they can shape interpretation without users realizing it.

Why AI still gets tripped up?

AI tools rely on pattern recognition, not true understanding. They predict what words should come next based on previous data. That means they often repeat outdated or biased information instead of adapting to current events.

The EBU and BBC study linked this to limited training data and cultural blind spots that make AIs miss nuance or context. Even when models seem intelligent, they are only reflecting what they have already seen, which makes them prone to spreading inaccuracies.

Developers push for better accuracy

OpenAI, Google, and Microsoft have all acknowledged their systems’ tendency to hallucinate or misquote. The EBU and BBC responded by releasing the News Integrity in AI Assistants Toolkit, a guide designed to help developers build more transparent and self-aware systems.

The toolkit focuses on teaching models when to stop guessing and how to cite real sources. Although these efforts are promising, experts agree that AI remains far from being consistently reliable in reporting or summarizing breaking news events.

How users can stay informed?

The EBU and BBC recommend that people treat AI-generated news as a starting point rather than a verified report. Always cross-check the information with credible journalists and established news outlets before sharing or believing it.

AI can summarize efficiently, but human judgment must come first. If something feels off, verify it manually. Using AI responsibly means applying skepticism, not blind trust. These habits help ensure that technology assists rather than replaces sound editorial judgment and factual awareness.

Signs an AI article may be unreliable

Experts say there are clear warning signs of unreliable AI news. Unnamed sources, broken links, vague attributions, and generic phrasing should all raise suspicion. The EBU and BBC study found that many AI-generated answers lacked identifiable evidence or clear sourcing.

If you cannot confirm where a claim originated, treat it as questionable. Developing these habits helps readers stay critical and prevents the spread of half-truths. In a world driven by automated content, verification remains a human responsibility.

The productive side of AI tools

AI remains valuable when used properly. Tools like ChatGPT, Copilot, and Paperpal can summarize reports, create outlines, and simplify technical documents. These uses carry less risk than relying on AI for breaking news accuracy.

The EBU and BBC suggest that AI works best as a helper for comprehension, not a substitute for journalists. When used thoughtfully, AI can save time and improve clarity while leaving the deeper investigation and fact-checking to trained professionals who ensure the truth.

How to use AI news wisely?

AI should be treated as a launch point for curiosity, not a source of final truth. Ask it to simplify complex topics, then check those facts through professional news organizations. Relying on multiple credible perspectives helps identify bias and false claims.

Experts emphasize that responsible readers blend AI convenience with critical thinking. Using AI wisely means knowing when to question, verify, and cross-reference information rather than taking every neatly phrased answer at face value.

The gap between expectation and responsible use is becoming clearer, as an MIT study finds AI failing at most companies that try to use it.

The takeaway for smarter readers

AI and journalism are evolving fast, but critical thinking remains constant. When a chatbot offers surprising or confident claims, pause to ask where they came from. Check the original source before sharing or reacting. The EBU and BBC study proves that speed can never replace accuracy.

Staying informed means valuing verification over convenience. In the race between automation and truth, thoughtful skepticism is the best defense for readers who want clarity in an AI-driven world.

That balance between curiosity and caution matters now more than ever, as OpenAI warns your chatbot might lie to you on purpose.

What do you think about this? Let us know in the comments, and don’t forget to leave a like.

Read More From This Brand:

• Meta Plans to Release a Standalone AI Chatbot
• Elon Musk’s AI chatbot sparks buzz with sudden political shift
• ChatGPT just went from chatbot to full-on project manager

Don’t forget to follow us for more exclusive content right here on MSN.

If you liked this story, you’ll LOVE our FREE emails. Join today and be the first to get stories like this one.

This slideshow was made with AI assistance and human editing.