AI chatbots distort news stories, BBC investigation reveals

Artificial Intelligence chatbots from major tech companies are struggling with accuracy when summarizing news articles, according to a comprehensive study by the BBC. The research evaluated the performance of ChatGPT, Google Gemini, Microsoft Copilot, and Perplexity across 100 BBC news articles to assess their ability to provide accurate news summaries.

Key findings: The BBC’s investigation revealed that more than half of all AI-generated summaries contained significant accuracy issues, with particular concerns about factual errors and quote manipulation.

• 51% of all AI responses contained major accuracy issues
• 19% of responses included incorrect statements, numbers, and dates
• 13% of quoted material was either altered or fabricated

Notable errors: Google’s Gemini emerged as the least reliable of the tested AI systems, with numerous instances of significant factual misrepresentation.

• Gemini incorrectly stated NHS policy on vaping, claiming the organization advised against it when they actually recommend it for smoking cessation
• ChatGPT made a serious error regarding Hamas leadership, referring to Ismail Haniyeh as alive months after his reported death
• Gemini’s responses were flagged for accuracy issues 46% of the time, the highest rate among tested systems

Industry precedent: Previous incidents have already prompted some tech companies to reconsider their AI-powered news features.

• Apple previously suspended its AI news summaries after the BBC identified inaccurate headline rewrites
• The company also modified its AI notification system to better distinguish AI-generated content from standard notifications

Call to action: BBC News leadership is urging for a collaborative approach to address these AI accuracy issues.

• BBC News CEO Deborah Turness expressed concerns about potential real-world harm from AI-distorted headlines
• There are calls for cooperation between news organizations, tech companies, and government bodies to address these challenges

Future implications: The widespread inaccuracies in AI-generated news summaries raise serious questions about the technology’s readiness for deployment in journalism applications. Until these accuracy issues are resolved, reliance on AI systems for news summarization could contribute to the spread of misinformation and potentially undermine public trust in both news media and artificial intelligence technologies.