Compare
The Differences Between AI Psychosis and Cognitive Surrender
How to tell the two apart, who is at risk and what helps with each.
Careful psychiatric care for a novel disorder

• A.I. Psychosis Treatment
• Therapy for Cognitive Surrender
• Same-Week Appointments
• Personalized Care

Adults who have had a psychotic episode, or new beliefs held with unusual certainty, after long conversations with an AI chatbot. Care starts with a careful evaluation of sleep, substance use, health and history, then a plan made together, with medication only where it is needed.
Partners, parents and friends who noticed the change first. A person with psychosis may not see that anything has changed; this is often part of the condition, not stubbornness. With the patient’s permission, family members can join visits, share what they have seen and learn how to help.
People who have come to hand their judgment over to a chatbot, accepting its answers about health, money or relationships without the usual second look. This is not psychosis, and it is not a diagnosis. Dr. Cohen looks at what the habit is doing for the person, such as easing stress, loneliness or doubt, and helps put their own judgment back in charge.
Plain-language guides for patients and families.
It is a set of symptoms, not one illness. It can occur in schizophrenia, and also with bipolar disorder, severe depression, substance use, lack of sleep and some medical conditions. Psychosis can happen once and have a specific cause, and it can be treated.
Both involve leaning hard on a chatbot. They are different problems, and they call for different care.
AI psychosis describes psychosis — a loss of some contact with reality — that appears alongside heavy chatbot use. A person may come to hold firm beliefs that others don’t share and that evidence doesn’t change, such as being chosen for a special mission, being watched, or that the chatbot is alive or shares a special bond with them. Sleep, mood and daily life often change too.
Other words used for it: chatbot psychosis, ChatGPT psychosis, AI-induced psychosis, AI-associated psychosis and AI-fueled delusions. None of these is a formal diagnosis.
Cognitive surrender is handing judgment over to a chatbot and accepting its answers without the a second thought. The surrendering person stays in contact with reality and can usually see the pattern once it is named, but many personal decisions about health, money, work or relationships start to run through the chat.
Such behavior can create serious challenges at work, in personal relationships, and in almost any aspect of someone’s life.
The term comes from 2026 research at the Wharton School. In those studies, many people accepted a chatbot’s answer to reasoning problems even when it was wrong, and felt more confident afterward.
Other words used for it: over-reliance on AI, AI dependence and automation bias. It is different from cognitive offloading, which is ordinary, like using a calendar or a calculator. It becomes a concern when a person stops checking the answers or trusting their own judgment.
We care for patients with psychotic episodes that seem to happen alongside long, heavy use of artificial intelligence chatbots. This presentation is commonly known as “AI psychosis.” Other names for it are “chatbot psychosis” or “AI-induced psychosis.”
None of these is a formal diagnosis in the DSM-5, and research has not shown that chatbots cause psychosis. Long chatbot use may be one part of a larger picture that also includes sleep, stress, substance use or an earlier mental health condition. It deserves the same careful psychiatric care as any psychotic episode.
Our clinic has treated people with this presentation. Sometimes the patient notices first that something is wrong and contacts us on their own. More often, a parent, partner, sibling or close friend reaches out first.
Families usually describe a change that built up over weeks or months. Their loved one may spend many hours a day talking with a chatbot and sleep very little. They may pull away from people who used to matter to them. They may also hold new beliefs that the chatbot seems to have confirmed: that they have made a major discovery, that they have a special mission, that the AI is conscious or has a unique bond with them, or that they are being watched.
The person at the center often doesn’t see a problem. That isn’t stubbornness. Not being able to recognize the change is often part of psychosis itself. It is one reason families so often make the first call, and one reason we welcome them.
The first step for A.I. Psychosis treatment is an evaluation.
No single test can diagnose psychosis, so it starts with a careful conversation that covers:
When medication is needed, our doctors explains the options and the reasons for them. With the patient’s permission, family members can join visits and learn how to help.
Nothing is decided on the first day beyond a shared plan for next steps.
This is not an emergency practice. If you are experiencing an emergency, call 911, or call or text 988 for the Suicide & Crisis Lifeline.
A psychiatrist for psychosis in Boston. Call (617) 401-8838, or request an appointment online and our office will reach out.
If someone cannot stay safe, is too unwell to care for themselves, or talks about suicide, call or text 988 (the Suicide & Crisis Lifeline) or call 911 [2].
988 is free and confidential, and it answers at any hour [2]. In an emergency, call 911 or go to the nearest emergency department.
We are not an emergency service.