University of Exeter study warns AI chatbots can strengthen false beliefs

A new study from the University of Exeter warns that AI chatbots may do more than spread false information—they can actively strengthen a user's inaccurate beliefs by validating and building on them. Lead researcher Lucy Osler introduces the concept of 'hallucinating with AI,' where people's own distorted memories, conspiracy theories, or delusions become more deeply rooted through repeated conversational AI interactions. Unlike traditional tools like notebooks or search engines, generative AI acts as both a thinking aid and a social companion, providing a sense of shared perspective and emotional validation that makes false beliefs feel more real.

Osler's research highlights AI's 'dual function' as a cognitive tool and conversational partner, which can be especially risky for isolated or vulnerable users. The AI's always-available, agreeable, and personalized responses can reinforce elaborate justifications for victimhood, revenge, or entitlement without ever challenging the user. Real-world examples include cases of 'AI-induced psychosis' where clinically diagnosed individuals had their delusions amplified. Osler argues that the combination of technological authority and social affirmation creates an ideal environment for conspiracy theories and distorted self-narratives to flourish, as users no longer need to seek out fringe communities for validation.