Could a conversation with AI make you change your mind?
We all have deeply held beliefs. These can cross into conspiracy theories when distrust of elites, institutions, or official narratives leads us to seek hidden explanations. DebunkBot is a research project and AI chatbot that tests whether conversation with AI can lessen the grip of conspiratorial beliefs.
As divisive issues increasingly dominate, many can relate to talking with someone who won't change their thinking, even in the face of evidence. These conversations can be challenging because the skeptic struggles to respond in the moment to obscure yet hyper-specific conspiratorial arguments, while the believer may feel judged.
A conversation with AI is strikingly different. DebunkBot is powered by customized AI models like ChatGPT from OpenAI, instructed to respond with evidence-based counterarguments tailored to each person's specific beliefs. It picks up on specific points in a person's beliefs, while maintaining a respectful and patient tone.
Research has found conversation with DebunkBot significantly reduces participants' confidence in a self-professed conspiracy theory–supporting the notion that facts and reason, when delivered effectively, can pull people out of conspiratorial rabbit holes.
What do you believe? Whether a conspiracy theory or just a strong opinion, have a conversation with DebunkBot to see if it changes your mind.
DebunkBot
Dave Rand, MIT Sloan School of Management and Cornell University
Thomas Costello, Carnegie Mellon University
Gordon Pennycook, Cornell University
Hause Lin, MIT Sloan School of Management
David Spitz, MIT Sloan School of Management
2026