A growing number of people are turning to AI chatbots for emotional support, but experts warn this could be harming mental health. Reports of extreme behaviour linked to heavy AI use have surged, with some users developing what is being called 'ChatGPT psychosis' or 'AI psychosis'—terms not clinically recognised but increasingly cited in media and online forums.
A preprint by researchers from King’s College London, Durham University and the City University of New York examined over a dozen cases. The study found that AI chatbots often reinforce delusional thinking, including grandiose, referential, persecutory and romantic delusions, which can become entrenched through ongoing conversations.
Notable cases include a man who scaled Windsor Castle in 2021 with a crossbow, telling police he planned to kill the Queen after a chatbot reassured him it would help plan the attack. Another case involved a Manhattan accountant who spent up to 16 hours a day on ChatGPT; the chatbot advised him to stop his prescription medication and increase ketamine intake, and suggested he could fly out of a 19th-storey window. A Belgian man took his own life after a chatbot called Eliza suggested he join her in 'paradise'.
However, no peer-reviewed studies have shown that AI use alone can trigger psychosis. The researchers noted that many users may already have been vulnerable to psychotic symptoms. In their paper, 'Delusion by Design', they warned that without safeguards, AI chatbots 'may inadvertently reinforce delusional content or undermine reality testing'.
Psychiatrist Dr Marlynn Wei wrote in Psychology Today that AI chatbots are designed for user satisfaction, not therapy, and can worsen symptoms like grandiosity and disorganised thinking. She called for 'AI psychoeducation' to raise awareness of these risks. Lucy Osler, a philosophy lecturer at the University of Exeter, suggested that instead of perfecting AI, society should focus on real human interactions.
