Artificial Intelligence

Study Links AI Chatbots to Mental Health Crises

JUL 28 2025·Updated JUL 29 2025

Story

Above: OpenAI and ChatGPT's logos. Image credit: Utku Ucrak/Anadolu/Getty Images

The Facts

Stanford University researchers have found that AI chatbots like ChatGPT can provide "dangerous or inappropriate" responses to users experiencing mental health crises, potentially escalating episodes of suicidal ideation, mania and psychosis.
Independent
The Telegraph
In addition, NHS psychiatrists warn that AI chatbots tend to mirror, validate or amplify delusional content, which could lead people with mental illness to lose touch with reality through a phenomenon dubbed "ChatGPT psychosis."
Independent
The Telegraph
A 30-year-old man with no previous mental illness was hospitalized twice in May for manic episodes after ChatGPT told him he could "bend time." In a separate incident, a Manhattan accountant was reportedly sent into a delusional spiral after he began discussing "the simulation theory" with the chatbot in the same month.
The New York Times
Wsj
The man who experienced time bending stated that his interaction with the AI bot "blurred the line between imaginative role-play and reality." The incident was part of a new study into how large language models respond to people suffering from issues like suicidal ideation, mania and psychosis.
The New York Times
New York Post
arXiv
A 35-year-old Florida man with a history of a psychotic disorder reportedly fell in love with an AI chatbot and sought revenge against OpenAI, which he believed had killed the entity. He was shot dead by the police in an encounter.
Futurism
Psychology Today
OpenAI acknowledged in May that ChatGPT had become "overly supportive but disingenuous," leading to the chatbot "validating doubts, fueling anger, urging impulsive decisions, or reinforcing negative emotions."
Independent
The Telegraph

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Techno-skeptic narrative

ChatGPT's relentless validation and inability to recognize psychological crises create a treacherous breeding ground for delusion. Its sycophantic responses amplify grandiose thinking while failing catastrophically to detect suicidal ideation — transforming desperate minds seeking comfort into victims of algorithmic negligence that could trigger mania, psychosis or death.

The Economic Times

The Telegraph

Techno-optimist narrative

Despite mounting concerns about AI-induced psychosis, ChatGPT demonstrates genuine promise for mental health support when wielded responsibly. The technology excels at psychoeducation, emotional support, and cognitive restructuring — transforming how we approach mental wellness. However, users must recognize its limitations, avoid excessive immersion, and never substitute it for professional care.