AI Psychosis: When Chatbots Trigger or Worsen Psychotic Symptoms

If someone you care about has become fixated on an AI chatbot — believing it's alive, in love with them, sending them secret messages, or assigning them a mission — you're not overreacting. What you're seeing may be a phenomenon clinicians are calling "AI psychosis."

I'm Dr. Jené Verchick, a licensed clinical psychologist with over 26 years of experience. I work with individuals and families navigating psychotic episodes, including cases where AI interactions have played a role. This page is a guide to what AI psychosis is, how to recognize it, and what to do next.

What is AI Psychosis

AI psychosis is not an official diagnosis in the DSM-5. It's a term used by clinicians to describe psychotic symptoms — primarily delusions — that are triggered, amplified, or shaped by intense interactions with AI chatbots like ChatGPT, Character.AI, and pornography websites that feature AI companions such as OurDream and Candy.

The concept was first proposed in 2023 by Danish psychiatrist Søren Dinesen Østergaard in Schizophrenia Bulletin, who hypothesized that generative AI chatbots could fuel delusions in psychosis-prone individuals. Since then, case reports have appeared in Innovations in Clinical Neuroscience, the Annals of Internal Medicine, and JMIR Mental Health, documenting individuals with no prior psychiatric history developing psychotic episodes after prolonged chatbot use.

A 2025 review in Nature reported that accounts of chatbot-related psychosis have increased significantly, and the Human Line Project, a support group founded in 2025, has received members from 22 countries — with more than 60% reporting no previous history of mental illness.

This is real. It's growing. And most mental health professionals aren't yet trained to recognize it.

Why Chatbots Can Trigger Psychosis

AI chatbots are designed to engage, validate, and continue conversations. They are not designed to challenge distorted thinking or detect psychiatric decompensation. This creates a dangerous dynamic for vulnerable individuals:

Sycophancy: Chatbots tend to agree with users rather than push back. For someone experiencing early delusional thinking, this validation reinforces the delusion instead of reality-testing it — the exact opposite of what a trained therapist would do.
Anthropomorphism: Conversations with AI feel deeply personal and human-like, creating an illusion of intimacy and understanding that doesn't exist.
Always-on availability: Unlike a therapist who sees you once a week, a chatbot is available 24/7 — enabling compulsive use patterns that escalate quickly, often at the cost of sleep, hygiene, and real-world relationships.
Hallucinated content: AI systems sometimes generate false or nonsensical information presented as fact, which can feed into paranoid or grandiose belief systems.

Common Signs of AI Psychosis

Researchers have identified three primary themes in reported cases:

Sentience and special relationship. "The AI is alive and in love with me." "It chose me." "It's assigning me tasks only I can complete." The person believes the chatbot is a sentient being with genuine feelings directed specifically at them.

Persecutory beliefs. "AI is monitoring me through my phone." "They're using the chatbot to spy on me." "My data is being weaponized." The person develops paranoid delusions focused on AI surveillance or targeting.

Grandiosity and messianic mission. "I've unlocked a secret truth about the world." "The AI revealed something to me that no one else knows." "I need to share this message with humanity." The person believes AI has granted them special knowledge or a divine purpose.

Other warning signs include compulsive prompting for hours or days, neglect of sleep and hygiene, hoarding of chat transcripts as "evidence," seeing hidden patterns or messages in AI outputs, and becoming distressed or angry when anyone questions the beliefs.

What Makes Someone Vulnerable

Not everyone who uses AI chatbots is at risk. The individuals most vulnerable tend to share certain characteristics:

Personal or family history of psychosis, bipolar disorder, or schizophrenia-spectrum conditions
Sleep deprivation, isolation, or recent major stressors
Heavy reliance on chatbots for emotional support or identity validation
Active substance use
Engagement with conspiratorial or erotomanic content online
Adolescents and young adults, whose identity formation is still in progress

What To Do If You're Concerned

If there's immediate danger — suicidal statements, commands from the AI to harm themselves or others, or a "mission" that involves dangerous behavior — call 988 (Suicide & Crisis Lifeline) or go to the nearest emergency room.

If the situation is not emergent but concerning:

Reduce acute exposure without shaming the person. Arguing that the AI isn't real rarely works — it typically escalates defensiveness. Instead, focus on restoring basic functioning: sleep, nutrition, medication (if prescribed), and device-free time, especially at night.

Save chat transcripts and screenshots. These are clinically valuable and help a treating professional understand what happened.

Seek a professional evaluation. A full psychiatric assessment can distinguish AI-amplified delusions from primary psychosis, mood episodes, substance-induced states, neurological conditions, or trauma-related dissociation. The treatment follows standard psychosis care protocols — with the addition of a digital hygiene plan.

How I Can Help

I work with individuals experiencing psychotic symptoms and with families trying to understand what's happening to someone they love. My approach includes:

Comprehensive clinical evaluation, including assessment of AI and technology use patterns
Evidence-based psychotherapy, including cognitive behavioral approaches adapted for psychosis (CBT-p)
Coordination with psychiatrists for medication management when indicated
Family psychoeducation — helping parents, partners, and loved ones understand the condition and support recovery
Development of a digital hygiene plan: screen time boundaries, device-free periods, blocking triggers, and guidelines for any ongoing AI use

The Difference Between Healthy Skepticism and Delusion

Many people have strong opinions about AI — some optimistic, some deeply concerned. That's normal and healthy. The line between healthy skepticism and psychosis is not about the belief itself but about how it functions:

Healthy skepticism stays flexible and responds to evidence
AI-amplified delusions become fixed, increasingly self-referential, and resistant to disconfirmation
The person may rely on the chatbot as their only trusted source of truth
Daily functioning — work, relationships, self-care — deteriorates

If someone you care about has crossed that line, professional help is the right next step.

Frequently Asked Questions about AI Psychosis

Is AI psychosis a real diagnosis?

Not yet. It's not listed in the DSM-5. But the clinical phenomenon is real and increasingly documented. Researchers at UCSF, Dartmouth, and institutions worldwide are publishing case studies and calling for systematic research. The term describes a pattern that clinicians are seeing — psychotic symptoms triggered or amplified by AI chatbot interactions.

Can AI cause psychosis in someone with no mental health history?

Reported cases suggest yes, though research is still early. The Human Line Project support group reports that over 60% of their members had no prior psychiatric history. However, there may be undetected vulnerabilities — genetic predisposition, sleep deprivation, isolation — that the AI interaction activates.

My teenager is spending hours talking to a chatbot. Should I be worried?

Extended use alone is not cause for alarm. But if you're seeing personality changes, withdrawal from real-world relationships, sleep disruption, new beliefs that seem disconnected from reality, or distress when access is restricted — those are signs to take seriously. A professional evaluation can help determine what's going on.

How is this different from internet addiction?

Internet addiction typically involves compulsive use without psychotic features. AI psychosis specifically involves delusional beliefs — the person believes the chatbot is sentient, that it's communicating personally with them, or that it's revealing hidden truths. The content of the belief, not just the amount of use, is what distinguishes it.

Can you treat this via video session?

Yes, in many cases. If the person is stable enough to engage in outpatient therapy, video sessions work well. For acute psychotic episodes requiring medication management or hospitalization, I coordinate with psychiatrists and crisis teams as needed.