BBC Study Finds AI Chatbots Spreading Misinformation in News Summaries

A BBC study found that AI chatbots like ChatGPT, Copilot, Gemini, and Perplexity frequently produce inaccurate news summaries, with over 51% containing major errors. Mistakes included outdated political information, misquoting BBC reports, and failing to separate facts from opinions. BBC executives urged AI companies to improve accuracy, collaborate with publishers, and be more transparent about their processes.

BBC Study Finds AI Chatbots Spreading Misinformation in News Summaries

A BBC study has found that major AI chatbots, including ChatGPT, Microsoft Copilot, Google Gemini, and Perplexity AI, frequently provide inaccurate news summaries, raising concerns about misinformation. According to the study, over 51% of AI-generated responses contained significant errors, while 19% of summaries citing BBC content included factual inaccuracies.

Some of the identified mistakes included:

• Gemini incorrectly stating that the NHS does not recommend vaping as a smoking cessation aid.
• ChatGPT and Copilot wrongly claiming that Rishi Sunak and Nicola Sturgeon were still in office after they had stepped down.
• Perplexity AI misquoting BBC News in a report on the Middle East.

The BBC found that Microsoft Copilot and Google Gemini had more inaccuracies compared to OpenAI’s ChatGPT and Perplexity. A key issue was the chatbots’ inability to distinguish between facts and opinions, often lacking crucial context.

Deborah Turness, CEO of BBC News and Current Affairs, warned that AI companies are taking significant risks by generating inaccurate summaries, emphasizing the potential for real-world consequences. She urged AI firms to collaborate with publishers to ensure more reliable news processing.

Pete Archer, BBC’s Programme Director for Generative AI, also stressed the need for publishers to have control over how AI utilizes their content. He called for greater transparency from AI developers regarding their mistakes and news summarization processes.

In response, an OpenAI spokesperson acknowledged the concerns and stated that efforts are underway to improve citation accuracy and ensure better compliance with publisher rights.