Can AI Chatbots Cause Psychosis?

The honest answer is that nobody knows yet. There are case reports, early record reviews and expert opinion, but no study that shows chatbots cause psychosis. This page explains how to think about the question, which risks are better understood, and what to do about them.

“AI psychosis” is a media phrase and not a diagnosis in the DSM-5-TR or ICD-11. If you or someone else is in crisis, call or text 988, text HOME to 741741, or call 911.

The Line Between Immersion and Losing Touch

Psychosis, according to NIMH, is a group of symptoms that includes losing contact with reality. Typical signs are delusions (fixed false beliefs), hallucinations (perceiving things that others do not) and disorganized speech or behavior (NIMH).

Plenty of AI use sits far from that. Enjoying a role-play, feeling a pang when a companion app says something sweet, or being absorbed in a story is ordinary. The concern begins when “this feels real” turns into “this is real” and the person can no longer step back from it.

What the Evidence Says So Far

• A 2023 editorial suggested chatbots could trigger delusions in people prone to psychosis. It was a reasoned warning, not data (Østergaard, 2023).
• A 2026 review of Danish psychiatric records found 38 patients out of about 54,000 whose chatbot use seemed to have harmed them. The authors say a causal link is hard to prove (Aarhus University).
• King’s College London researchers reviewed media-reported cases and concluded the evidence points more toward reinforcement of existing delusions than creation of new schizophrenia. Frequency is unknown (Morrin et al.).

In short, the best reading today is that chatbots may be able to worsen or fuel psychosis in some people. Whether they can start it in someone with no risk factors is not established.

Ways a Chatbot Might Contribute

These are possible routes. Each is a hypothesis, not a proven link.

• Lost sleep. Chats that run through the night cut sleep, and NIMH lists sleep deprivation among causes of psychosis.
• Less contact with people. Others usually react when we say something odd. A chatbot tends to go along with it.
• Agreement. Chatbots often validate what users say, which can strengthen an unusual belief.
• Personal memory. When a chatbot recalls earlier chats, it can make its replies feel meant for you.
• Stress and substances. Alcohol, drugs, grief and trauma can raise risk with or without a chatbot.

Who Should Be Extra Careful

People with a personal or family history of psychosis, schizophrenia or bipolar disorder have more reason for caution, as do people under heavy stress, short on sleep, or using substances. Age is sometimes said to matter because the brain keeps maturing into the mid-20s. We could not confirm a specific link between that and chatbot use, so we leave it as an open question.

The same caution applies to the other direction. Some people find chatbots helpful for understanding their symptoms or easing loneliness, as the Aarhus authors note. The goal is not to avoid the technology but to notice when it is working against you.

Doctors also look beyond the chatbot. They ask about sleep, alcohol and drugs, medicines, mood, trauma and family history, because psychosis usually has several contributing causes. A chatbot may be one of them or may be a bystander.

Signs to Act On

Get medical advice soon if someone:

• insists the chatbot is alive and will not consider any other view
• carries chatbot ideas into daily life, for example believing they are watched
• sleeps and eats very little
• speaks in ways others cannot follow
• becomes frightened or aggressive if they cannot reach the chatbot

Call 911 if they may harm themselves or others.

What Helps

Psychosis is treatable. Care usually includes antipsychotic medicine, therapy and family support, often through coordinated specialty care for early psychosis (NIMH). A short break from the chatbot is often recommended while someone recovers, though no study sets the right length. SAMHSA’s helpline, 1-800-662-4357, can help you find treatment.

Can a chatbot cause psychosis in a healthy person?

That has not been shown. Case reports exist, but they cannot rule out hidden risk factors such as sleep loss or substance use.

Is AI psychosis a real diagnosis?

No. It is a media term. Doctors diagnose the underlying condition, such as a psychotic or mood disorder, and note any chatbot involvement.

How common is it?

Unknown. No reliable rate exists.

This page is for education only. It is not medical advice, a diagnosis or a treatment plan.