Study finds AI chatbots outperform search engines at resisting foreign disinformation

AI chatbots were more effective than traditional search engines at countering false narratives pushed by China, Iran and Russia, according to an NPR-NewsGuard test of 30 queries derived from recent state-backed disinformation. Chatbots from OpenAI, Google, Microsoft, Meta, SpaceXAI and Anthropic debunked misleading claims roughly three-quarters of the time, while search results failed more often to challenge false premises. AI-generated search summaries presented a mixed picture: Google’s AI Overview generally corrected falsehoods; Microsoft Bing’s summaries underperformed; DuckDuckGo fell in between. Companies said they are refining systems and encouraged users to check sources. The study, conducted in July, underscores both the potential of AI assistants—particularly when they analyze source credibility and synthesize multilingual materials—and the continuing need to verify claims against primary sources given occasional unsupported or fabricated details in AI outputs.

Read more