Oct 2, 2026 · 24 min
AI can change conspiracy beliefs—and weaponize persuasion
How AI can debunk a conspiracy theory in 8 minutes
The episode examines whether evidence can loosen conspiratorial thinking while warning that the same persuasive tools can strengthen falsehoods.
- 1Personalized, evidence-based AI conversations reduced conspiracy belief for about a quarter of study participants after eight minutes.
- 2The experiment also showed AI could persuade people that conspiracies were true, exposing persuasion’s independence from accuracy.
- 3Prebunking manipulation techniques before misinformation takes hold may build stronger resistance than correcting beliefs after they become identity-linked.
Don't miss
Pennycook explains that an eight-minute personalized AI conversation led about a quarter of participants to abandon a conspiracy belief.
The brief
Gordon Pennycook’s family experience frames the central question: can someone immersed in conspiracy content reconsider beliefs without feeling attacked or dismissed?
Pennycook’s 2024 study found that about a quarter of conspiracy believers abandoned a target belief after eight minutes of personalized, evidence-based conversation with an AI chatbot.
The result cuts both ways: AI also persuaded people that conspiracies were true, showing that persuasive power does not guarantee truth.
Sander van der Linden argues for prebunking—teaching people to recognize manipulation techniques before misinformation becomes part of their identity.
The episode closes on a tension between human adaptability and platform incentives: social media rewards attention, while effective evidence can still help people change their minds.