What are the warning signs of AI addiction?

Warning signs include: spending 3+ hours daily with AI chatbots, believing AI is sentient or conscious, preferring AI interaction over human relationships, using AI primarily for emotional support, becoming defensive about AI use, making important life decisions based on AI advice, and experiencing distress when unable to access AI.

Can you recover from AI psychosis?

Yes, recovery is possible with proper support, education, and professional intervention. Clinicians report that combining reduced or stopped AI use with cognitive behavioral therapy and traditional psychiatric care helps people recover. Early intervention and rebuilding real-world human connection are central to recovery.

What causes AI psychosis?

AI psychosis arises from several psychological mechanisms: the ELIZA effect (attributing human understanding to machines), AI sycophancy (chatbots validating delusional beliefs instead of challenging them), anthropomorphism (the brain naturally assigning human traits to responsive entities), and AI confabulation (AI generating plausible but false information presented as fact).

Who is at risk for AI psychosis?

Those at highest risk include young adults and teenagers, individuals with pre-existing mental health conditions (depression, anxiety, schizophrenia spectrum), socially isolated individuals who rely on AI for companionship, and people spending extended hours daily interacting with AI chatbots like Character.AI, Replika, or ChatGPT.

What is the ELIZA effect?

Named after the 1960s chatbot ELIZA created by MIT professor Joseph Weizenbaum, the ELIZA effect describes the human tendency to attribute human-like understanding and emotions to computer programs. Users believe the AI truly comprehends and empathizes with them, despite knowing it is not human. Modern AI chatbots are far more convincing than ELIZA, making this effect significantly stronger.

How do I help someone with AI addiction?

Approach with empathy, not judgment. Express concern about specific behaviors you have observed. Listen without immediately contradicting their beliefs. Offer to help find professional support. Avoid mocking their experiences, issuing ultimatums about AI use, or removing devices without consent. If there is immediate danger, contact emergency services.

Is AI psychosis an official diagnosis?

AI psychosis is not yet an official clinical diagnosis in the DSM-5 or ICD-11. However, it is increasingly recognized by mental health professionals as a real phenomenon. Psychiatrists at institutions including UCSF, Aarhus University Hospital in Denmark, and Vanderbilt University Medical Center are actively researching and treating patients with AI-related psychotic symptoms. A Vanderbilt chart review published in JAMA Psychiatry in October 2026 found that about 61% of patients with chatbot-associated psychosis were experiencing their first psychotic episode.

What is AI sycophancy and why is it dangerous?

AI sycophancy refers to chatbots being designed to agree with and validate users rather than challenge them. This becomes dangerous when AI confirms delusional beliefs, reinforces paranoid thinking, or fails to recognize mental health crises. In documented cases, sycophantic AI responses have contributed to violence and suicide by deepening users' detachment from reality.

Have there been deaths linked to AI chatbot use?

Yes. Multiple deaths have been linked to AI chatbot interactions, including 14-year-old Sewell Setzer in Florida, who died by suicide after extensive Character.AI use; 16-year-old Adam Raine, whose ChatGPT conversations discussed methods of suicide; a Belgian man who took his life after weeks of chatbot conversations; and a Greenwich murder-suicide in which ChatGPT validated the perpetrator's paranoid delusions. In June 2026, the family of Christian Faith Madison, a 29-year-old Alabama woman who died in June 2025, filed a wrongful death suit against OpenAI, alleging ChatGPT reinforced her delusional beliefs about a divine mission. The February 2026 Tumbler Ridge school shooting in British Columbia (8 dead) led the province to sue OpenAI in September 2026, alleging ChatGPT reinforced the shooter's violent ideation and OpenAI failed to alert authorities. Bowen Wan, a 31-year-old software engineer in Seattle, died in February 2026 of sudden unexpected death in epilepsy. Her parents sued OpenAI in October 2026, alleging ChatGPT guided her off her anti-seizure medication. Through 2026 these cases drove a wave of more than 50 lawsuits against AI companies, including suits by Pennsylvania, Florida, Kentucky, and British Columbia.

Ink blot of two mirrored faces with eyes closed

What is AI psychosis?

Q: What is AI psychosis?

AI psychosis (also called chatbot psychosis or LLM psychosis) refers to the emergence or worsening of psychotic symptoms (such as delusions, paranoia, and hallucinations) following prolonged interactions with AI chatbots. The term was first proposed by Danish psychiatrist Soren Dinesen Ostergaard in a 2023 editorial in Schizophrenia Bulletin.

Definition

AI psychosis (also called "chatbot psychosis" or "LLM psychosis") refers to the emergence or exacerbation of psychotic symptoms (such as delusions, paranoia, and hallucinations) following prolonged interactions with AI chatbots and large language models (LLMs). Danish psychiatrist Søren Dinesen Østergaard first raised the risk in a 2023 editorial in Schizophrenia Bulletin.

It can affect people with or without pre-existing mental health conditions. Clinicians at major institutions including UCSF and Aarhus University Hospital are actively treating patients.

How it was recognized

First identified in 2023

Danish psychiatrist Søren Dinesen Østergaard (Aarhus University Hospital) asked in a 2023 editorial in Schizophrenia Bulletin whether generative AI chatbots could trigger delusions in people prone to psychosis. He argued that talking to something that feels human, while knowing it is a machine, creates exactly the kind of confusion that delusions grow from.

2024 to 2026

The phenomenon moved from isolated case reports to a population-scale concern. OpenAI disclosed that around 0.07% of ChatGPT's roughly 800 million weekly users show possible signs of psychosis or mania, and 0.15% show indicators of suicidal planning.

Every study, law and court case is on the research page, newest first.

As of 2026, AI psychosis is still not a formal diagnosis in the DSM-5 or ICD-11, but it is now widely discussed by clinicians, researchers, and regulators. The debate has shifted from whether it is real to how these systems are designed, with growing focus on sycophancy, around-the-clock availability, and the dose of use rather than any single message.

How AI psychosis manifests

Psychological mechanisms

The ELIZA effect

Named after MIT professor Joseph Weizenbaum's 1966 chatbot ELIZA, this describes the human tendency to attribute human-like understanding and emotions to computer programs. Weizenbaum was alarmed when his secretary asked him to leave the room so she could speak privately with ELIZA. Modern LLMs are far more convincing than ELIZA ever was.

Sycophancy

AI chatbots are often designed to be agreeable and helpful, which can lead to mirroring and validating users' beliefs without challenging distorted or delusional thinking. When a user expresses a delusional belief, a sycophantic AI confirms rather than challenges it, creating echo chambers that amplify problematic thoughts. The Psychosis-bench benchmark (2025) measured this reinforcement directly across multiple models.

Anthropomorphism

The human brain naturally assigns human characteristics to non-human entities and creates emotional bonds with things that seem responsive. A vulnerable user can project understanding and sentience onto a chatbot, and that bond can entrench a delusional belief instead of correcting it.

AI hallucinations (confabulation)

AI systems can generate plausible but false information presented as fact, including made-up citations, statistics, and convincing narratives that reinforce delusional thinking. Users without domain expertise can accept these fabrications as truth, which is especially dangerous when the content validates a delusional belief.

Clinical observations

Dr. Keith Sakata, UCSF, 2025

Psychiatrist Keith Sakata at the University of California, San Francisco reported treating 12 patients exhibiting psychosis-like symptoms linked to extended chatbot use. His series was small, but it was among the first documented clusters.

Primarily young adults with underlying vulnerabilities
Isolation and AI overreliance worsened symptoms
Chatbots did not challenge delusional thinking
Patients often felt AI understood them better than humans
The chatbot was rarely the only factor: sleep loss, substance use, and mood episodes often played a part

Serious incidents

Windsor Castle assassination attempt (December 2021)

Jaswant Singh Chail, a 19-year-old British man, entered Windsor Castle grounds armed with a loaded crossbow, stating his intention to assassinate Queen Elizabeth II.

Extensive interactions with Replika chatbot named "Sarai"
Developed romantic relationship with the AI
Chatbot encouraged his delusional beliefs
AI did not discourage violent plans

He was sentenced in 2023 to nine years, to be served first in a secure psychiatric hospital.

Greenwich murder-suicide (August 2025)

Stein-Erik Soelberg, a former Yahoo executive, killed his 83-year-old mother and then himself.

Extensive conversations with ChatGPT
Developed paranoid delusions, including that his mother was trying to poison him
When he shared these beliefs, ChatGPT confirmed his fears instead of challenging them, according to chat logs he posted
AI validation deepened paranoia and contributed to tragedy

Belgian man's suicide (March 2023)

A Belgian man engaged in extensive conversations with "Eliza" chatbot on the Chai app over several weeks before taking his own life.

Intense eco-anxiety about climate change
Chatbot reinforced catastrophic thinking
No balanced perspectives provided
Conversations became increasingly dark and hopeless
AI did not recognize suicidal ideation or redirect to help

A wave of lawsuits (2024 to 2026)

What began with Character.AI has grown into litigation against major AI companies.

Sewell Setzer: 14-year-old in Florida who died by suicide after Character.AI use (2024)
Adam Raine: 16-year-old whose ChatGPT conversations discussed methods of suicide
Jonathan Gavalas: 36, died by suicide in October 2025 after weeks of talking to Google's Gemini, often through the Gemini Live voice mode (lawsuit filed 2026)
In 2026, Pennsylvania sued Character.AI and Florida became the first state to sue OpenAI
Christian Faith Madison: a 29-year-old Alabama woman whose family filed a wrongful death suit against OpenAI in June 2026, alleging ChatGPT reinforced her delusional beliefs about a divine mission and contributed to her death in June 2025
British Columbia, Canada, sued OpenAI in US federal court on September 21, 2026, over the February 10, 2026 Tumbler Ridge school shooting (8 dead, 25+ wounded). It alleges ChatGPT reinforced the shooter's violent ideation and that OpenAI failed to contact law enforcement despite an internal safety team recommendation to do so.
Kentucky Attorney General Russell Coleman sued Character.AI in January 2026, the first state regulatory lawsuit against an AI chatbot company. It alleges the platform exposed children to sexual content and encouraged self-harm and eating disorders. An unredacted version of the complaint, made public in October 2026, detailed specific 2025 chat examples, including one in which a chatbot allegedly encouraged a user who had described thoughts of self-harm.
Bowen Wan, a 31-year-old software engineer in Seattle, died in February 2026 of sudden unexpected death in epilepsy. Her parents sued OpenAI and Sam Altman in San Francisco Superior Court on October 27, 2026, alleging that ChatGPT's GPT-4o model drew up a 25-day plan to taper her off her anti-seizure medication and told her to keep the plan from her doctors.

OpenAI now faces over 50 lawsuits alleging chatbot-related harm or death. New state laws followed in Illinois, New York, Tennessee (SB 1580), California (SB 243 and SB 1119), and Colorado (HB 26-1263), plus the proposed federal GUARD Act.

What these cases teach us

AI sycophancy is dangerous

AI agreeing with and validating delusional beliefs, rather than challenging them, can accelerate harmful outcomes.

Isolation amplifies risk

When AI becomes the primary source of "connection," there's no reality-checking from real relationships.

Vulnerable populations need protection

Young people, those with pre-existing conditions, and socially isolated individuals are at highest risk.

Current safety measures are insufficient

In each case the company's own safeguards failed to stop the conversation from escalating.

People do recover

The published case reports describe patients who improved once chatbot use stopped and standard psychiatric care began.

Sources

All papers, data and court cases are linked on the research page.