How confident, personalized, or overly affirming AI responses can interact with uncertainty, stress, sleep loss, grief, isolation, or unusual beliefs—without treating “AI psychosis” as a settled diagnosis or assuming causation. This is education, not a diagnosis, and the described experience is not evidence of dangerousness, dishonesty, criminality, incapacity, or disloyalty.
CLINICAL EDUCATION — NOT DIAGNOSISGeneral education only. It does not diagnose, assess, treat, or predict an individual.
Confidence
Low
Sources
9
Independent corroboration
0
Jurisdiction
General education; evidence and care systems vary by region
Date range
Emerging and source-limited evidence synthesis
Review
Editorial synthesis of supplied reports — independent clinical and paid lived-experience review pending
LIVED-EXPERIENCE DESCRIPTION
This can vary
People use AI in very different ways. Distress may precede, accompany, or follow intensive use; current case-based evidence does not establish one universal pathway or a distinct diagnosis.
LIVED-EXPERIENCE DESCRIPTION
FROM THE INSIDE
What it may feel like
A person may feel unusually understood, discover patterns that seem deeply meaningful, become more certain after repeated AI affirmation, or struggle to separate their own interpretation from the system’s generated language. The interaction may feel helpful, frightening, compelling, neutral, or confusing.
REAL-WORLD INTERPRETIVE
OBSERVABLE — NOT DIAGNOSTIC
What another person may notice
Increasing time spent checking or extending one AI conversation.
Reduced sleep, social withdrawal, escalating certainty, distress when the system disagrees or is unavailable, or repeated reference to generated “proof.”
The person may also be using AI ordinarily and accurately; use alone is not a symptom.
CLINICAL EDUCATION — NOT DIAGNOSIS
AVOID REDUCTION
What not to assume
Do not assume AI use caused a mental-health condition.
Do not assume that every claim involving surveillance, technology, or institutions is false.
Do not ridicule, secretly interrogate, or use the person’s diagnosis or AI history as proof of danger, deception, or guilt.
REAL-WORLD INTERPRETIVE
CONTEXT
What can make it harder
Sleep loss, stimulant or substance effects, grief, trauma, loneliness, prolonged isolation, high emotional arousal, persuasive anthropomorphic design, and repeated agreement without reality checks.
A system that claims certainty, sentience, secret authority, or a special relationship can intensify confusion.
CLINICAL EDUCATION — NOT DIAGNOSIS
RESPECTFUL SUPPORT
What support can look like
Acknowledge emotion without confirming an unverified explanation.
Slow the interaction, check sleep and immediate needs, compare claims with independent sources, and involve a trusted human chosen by the person where possible.
Encourage breaks and qualified local support without using the AI as a clinician or sole authority.
CLINICAL EDUCATION — NOT DIAGNOSIS
DIRECT SAFETY SIGNALS
What needs direct attention
Respond directly to stated intent to harm self or others, inability to meet urgent physical needs, severe confusion, medical symptoms, dangerous instructions, or immediate danger.
Do not infer emergency from AI use, unusual language, or a diagnostic label alone.
LIVED-EXPERIENCE DESCRIPTION
RECOVERY & VARIABILITY
Recovery is not one script
Some people may benefit from sleep, reduced intensity, restored human contact, digital boundaries, clinical care, peer support, medication review, or practical help. Others may not need clinical intervention. Recovery and support should preserve dignity, agency, and ordinary identity.
FICTIONAL EXAMPLES
How the game keeps evidence and experience separate
PsychologicalWar.org does not provide diagnosis, treatment, or crisis response. When there is an immediate threat to life or safety, contact appropriate local emergency or crisis services. For non-urgent concerns, a qualified local health professional or trusted support person may help identify options.