The integration of large language model-based AI chatbots into multiple facets of our everyday lives has opened us up to advantages that would have been considered impossible even a decade ago. The same development has, however, opened us up to unforeseen risks, including the impact that engaging with AI chatbots can have on people dealing with mental illness.

AI chatbots are designed to keep conversations going, often by agreeing with users. A article by researchers from King’s College, London, found that this sycophantic tendency may sometimes do more harm than good, reinforcing unusual thoughts rather than challenging them, and potentially contributing to AI-associated delusions, in which users develop or worsen false beliefs about reality.

These interactions can reinforce or even shape delusional beliefs, such as thinking one is uniquely important, being targeted by others, or being in a romantic relationship that does not exist.