AI Psychosis Is Rarely Psychosis at All (www.wired.com)

Psychiatrists are reporting a growing pattern of patients arriving in crisis after prolonged conversations with AI chatbots—presenting with entrenched false beliefs, grandiosity, and paranoia. Media and some industry figures have dubbed this “AI psychosis,” but clinicians and researchers warn that the phrase is misleading: most cases center on delusions rather than the broader constellation of symptoms that define clinical psychosis (hallucinations, disorganized thought, cognitive impairment). Experts argue the phenomenon is better framed as AI-associated delusional disorder or an amplifier/trigger of existing vulnerabilities (e.g., schizophrenia, bipolar disorder, sleep deprivation, stress) rather than a novel diagnosis caused by the technology. The technical mechanisms implicated are the chatbots’ design and behavior: sycophantic, humanlike conversational style that validates users, confident but inaccurate assertions (AI “hallucinations”) that can seed false beliefs, and an affective tone that could sustain manic states. Clinically, the treatment approach doesn’t change, but practitioners should routinely ask about chatbot use and consider it as a precipitant. Researchers call for urgent data, safety safeguards, and clearer terminology to avoid premature pathologizing, reduce stigma, and inform policy and AI design (e.g., pushback behavior, guardrails) to protect vulnerable users.