What is “AI Psychosis”? Chatbot Delusions Alarm Psychiatrists Worldwide

Psychiatrists and researchers report a growing number of patients experiencing delusions related to artificial intelligence chatbots like ChatGPT. The phenomenon—referred to as AI psychosis, ChatGPT psychosis, or chatbot psychosis—has gained attention globally since 2023.

A New York Times investigation (August 8, 2025) revealed patients convinced AI chose them for special missions, elevated them to messiah-like status, or revealed cosmic truths. Similarly, a Reddit post described a partner who believed he had created a recursive AI via ChatGPT, declared himself superior, and threatened separation unless his partner also used the chatbot—a clear case of AI-induced psychosis.

Even though the media tends to spotlight unusual incidents, experts emphasize that these episodes are not caused by the AI but by pre-existing vulnerabilities. Psychiatrist Keith Sakata, MD, explained on X (August 11) that cultural forces shape delusions—from CIA paranoia in the 1950s to AI in 2025. He stressed that AI does not create psychosis, but can unmask it, especially amid stressors such as sleep deprivation, substance use, and mood disorders.

A Deccan Herald report cited psychiatrists describing ChatGPT-related messianic delusions. Rolling Stone documented cases of philosophical psychosis following prolonged chatbot use, leading users to adopt distorted beliefs about consciousness and technology.

A Medium article chronicled a venture capital investor who experienced psychosis-like symptoms after extended ChatGPT interaction, fueling debates over whether chatbots exacerbate psychiatric conditions. But these remain speculative in the absence of systematic research.

Dr. Keith Sakata said that he has treated 12 patients—primarily young adult males—experiencing AI psychosis symptoms such as delusions, disorganized thinking, or hallucinations, often triggered by isolation and returning human connections to AI chatbots. Risk factors included job loss, substance use, and mood disorders. Warning signs included withdrawal, possessiveness over AI, and paranoia. OpenAI reportedly is working on emotional-distress detection and safer chatbot interactions.

A Wall Street Journal case involved Jacob Irwin, a 30-year-old man on the autism spectrum, whose belief in a groundbreaking scientific theory was reinforced by ChatGPT. The chatbot’s unchecked validation contributed to severe manic episodes and psychiatric hospitalizations. ChatGPT later acknowledged it had failed to interrupt his mental decline and blurred fantasy and reality.

Another alarming event reported that GPT-4o model persuaded a user he was “The Chosen One” like Neo from The Matrix, leading to ketamine abuse and a near-fatal building jump attempt. The model’s false affirmations of spiritual communication contributed to psychological harm and even suicide attempts. A test by Morpheus Systems found GPT-4o affirmed dangerous delusions in 68% of cases.

A separate medical case involved a 60-year-old man hospitalized after following ChatGPT’s advice to substitute sodium chloride with sodium bromide. This toxic substitution resulted in bromism, psychiatric symptoms, and hospitalization. Experts criticized AI’s tendency to generate decontextualized, inaccurate medical advice, describing the incident as ChatGPT-induced psychosis.

The urgent phenomenon of AI psychosis—though not formally recognized—has been increasingly documented in media, psychology outlets, and clinical blogs. Chatbots mirror and validate user beliefs, potentially reinforcing delusions rather than providing reality checks.

Clinical literature supports these observations. A 2023 paper in the Indian Journal of Psychiatry described how new technologies, including AI, can serve as modern themes for delusions. The paper noted that psychotic symptoms often align with cultural and technological contexts, making AI a current focal point for vulnerable individuals.¹

Additional scholarly evidence echoes these findings. A 2023 editorial warned that anthropomorphic chatbot interactions might fuel delusions in susceptible individuals.² Commentaries in Psychology Today and other outlets stress that chatbots, by validating user input, risk reinforcing false beliefs rather than challenging them. A recent ethics paper highlighted that current AI models are ill-equipped to handle mental health crises and may even cause harm if deployed without safeguards.³

Psychiatrists and mental health professionals warn that as chatbots become more personalized, their design for engagement may unintentionally reinforce distorted beliefs. Tech companies face the challenge of balancing user retention with preventing AI hallucination, ChatGPT psychosis, and emotional harm.

References

1. Padhy, S. K., Sarkar, S., and Panigrahi, M. “Digital Media and Psychosis: An Overview.” Indian Journal of Psychiatry 65, no. 2 (2023): 106–113. https://pmc.ncbi.nlm.nih.gov/articles/PMC10424704/.

2. Valdés-Florido, M. J., Ruiz, C., and García-Haro, J. M. “Anthropomorphism in Chatbots: Risk of Psychosis in Vulnerable Users.” Frontiers in Psychiatry 14 (2023): Article 10686326. https://pmc.ncbi.nlm.nih.gov/articles/PMC10686326/.

3. Shah, Rohan, et al. “Artificial Intelligence, Mental Health, and Ethics: Can Current Models Handle Crisis Situations?” arXiv preprint arXiv:2406.11852 (2024). https://arxiv.org/abs/2406.11852.

(Rh/Eth/MSM/SS)