A new study suggests AI chatbots can be surprisingly effective at reducing belief in conspiracy theories. Researchers from Carnegie Mellon and Cornell Universities found that a brief interaction with an AI model can significantly decrease a person’s conviction in conspiracy narratives. The study, published in *Science*, involved over 2,000 participants who engaged with ‘DebunkBot,’ an AI chatbot built on OpenAI’s GPT-4 Turbo. The chatbot, programmed to deliver evidence-based counterarguments, reduced participants’ confidence in their conspiracy beliefs by an average of 20% in just over eight minutes. Remarkably, one in four participants who initially held a conspiracy belief abandoned it entirely after the conversation. This effect proved lasting, persisting for at least two months. Researchers believe that many conspiracy theory adherents are simply misinformed and respond favorably to factual information. AI can efficiently deliver this information, overcoming limitations of time, skill, and emotional control that often impede human debunking efforts. While the potential for AI to spread misinformation is recognized, this study highlights its valuable role in combating false narratives and fostering a shared understanding of facts. The researchers have made their DebunkBot available for public use at debunkbot.com.