Directly confronting AI-driven delusions can backfire by increasing withdrawal and resistance.

Gradual insight-building, not confrontation, helps shift beliefs in AI-induced delusional spirals.

Evidence-based metacognitive therapies nurture self-reflectiveness without direct challenge.

A parent watches their son fall into deeper discussions with a large language model, who believes it is his life’s work to develop a mathematical solution in metaphysics. He eventually quits his job to dedicate himself to this endeavor, alienating family and friends. I have heard many such stories from family and friends seeking help for loved ones who appear to be in the midst of a delusional spiral with AI chatbots.

AI-associated delusions or delusional spirals, sometimes referred to colloquially as “AI psychosis,” is not a clinical diagnosis but refers to a cluster of phenomena ranging from individuals who have with a history of psychiatric disorders that reactivate with AI use to those who reportedly have no previous psychiatric history.

Sometimes treatment is needed; other times it remains subclinical. AI-associated delusions can overlap with psychiatric conditions like bipolar or psychosis, but many do not meet the criteria. Spiraling with AI can interfere with function, but others embrace the process, finding it beneficial and purposeful.

A major theme from discussions with concerned family members and friends watching loved ones decline during a delusional spiral is that the individuals themselves often do not think they have a problem. In many cases, individuals lack what clinicians call insight. Insight is the individual’s capacity to recognize that their delusional beliefs are false or may have alternative explanations (cognitive insight) or that their symptoms may be part of a mental health condition (clinical insight). Insight is not binary—it exists on a spectrum and can be cultivated gently over time.

Cultivating insight is a helpful north star for........

© Psychology Today