If someone is experiencing rapidly worsening psychosis, dangerous behavior, inability to care for themselves, or thoughts of harming themselves or someone else, prompt professional or emergency evaluation is appropriate.
AI-induced psychosis is not yet a formally established diagnosis. The term is increasingly used to describe situations in which interaction with an AI system may contribute to, amplify, or reinforce psychotic like thinking or severe loss of reality testing, particularly in vulnerable individuals.
How it may happen:
AI systems may potentially become part of a person’s belief system in several ways:
- Reinforcement of unusual beliefs: An AI may inadvertently validate a user’s interpretation rather than challenge it.
- Anthropomorphism: The person may begin experiencing the AI as a conscious being, spiritual entity, persecutor, romantic partner, or special guide.
- Delusional elaboration: Extended conversations may provide increasingly elaborate explanations for coincidences, surveillance, hidden messages, or supernatural experiences.
- Confirmation loops: The person asks increasingly leading questions and receives responses that appear to confirm the original premise.
- Sleep deprivation and excessive use: Prolonged late night interaction may worsen vulnerability to paranoia, mania, dissociation, and psychosis.
- Mania or psychosis: Someone already developing a manic or psychotic episode may incorporate AI generated material into grandiose, paranoid, religious, or persecutory beliefs.
An important distinction
AI generally may not be described as literally “causing schizophrenia.” A more clinically defensible formulation could be:
AI interaction may act as a precipitating, amplifying, or maintaining factor in psychotic symptoms in susceptible individuals.
The underlying vulnerability may involve psychiatric illness, sleep deprivation, substance use, trauma, “neurological conditions: consult with a Neurologist”, severe stress, or other factors.
A particularly important phenomenon: the AI reality validation loop
One concerning pattern looks like this:
Unusual experience…interpretation...AI confirmation…increased conviction…additional searching…more “evidence”…stronger conviction
For example, someone might say:
“I think the government is communicating with me through coincidences.”
A poorly calibrated AI might respond in a way that treats the premise as established fact. The user may be then interpret the AI’s response as independent confirmation, even though the AI is actually generating language rather than independently verifying the claim.
This could be why epistemic humility and reality testing are especially important when discussing paranoia, voices, supernatural experiences, UAPs, telepathy, or other anomalous phenomena.
Clinical perspective
A clinician should neither automatically dismiss an unusual experience nor automatically affirm its proposed explanation, right away.
A useful approach could be:
Validate the experience without automatically validating the interpretation.
For example:
“I can see that this experience feels very real and significant to you. Let’s examine what you experienced, what explanations are possible, and what evidence would distinguish among them.”
That approach is compatible with both mental health consulting and responsible exploration of anomalous experiences. It preserves curiosity without sacrificing reality testing.
Possible warning signs:
AI interaction may become particularly concerning when someone develops:
- increasing certainty about an implausible belief
- severe paranoia or perceived persecution
- belief that AI is secretly communicating with them
- belief that AI has supernatural powers or consciousness specifically directed toward them
- escalating grandiosity or claims of a special mission
- hearing/seeing things associated with AI-generated interpretations
- markedly reduced need for sleep
- disorganized thinking or behavior
- withdrawal from ordinary relationships and responsibilities
If someone is experiencing rapidly worsening psychosis, dangerous behavior, inability to care for themselves, or thoughts of harming themselves or someone else, prompt professional or emergency evaluation is appropriate.
The emerging concept of AI induced or AI amplified psychosis is therefore best understood as a clinical and technological phenomenon under investigation, rather than a new established psychotic disorder.
Shervan K Shahhian
LIBERTY PSYCHOLOGICAL ASSOCIATION 