Information only, not medical advice or a diagnosis. In a crisis in the US, call or text 988.

Artificial Intelligence Use Disorder (AIUD): A DSM-5-TR Style Diagnostic Classification

First published September 2025. Copyright: The AI Addiction Center

This is not a diagnosis, and AIUD is not an official condition. It is an unofficial working draft written by the editors of The AI Addiction Center, with no clinician or journal behind it. AI chatbot addiction is not in the DSM-5-TR or ICD-11. Do not use this page to diagnose yourself or anyone else. If you are worried, see a licensed professional. In crisis, call or text 988, or text HOME to 741741 (Crisis Text Line).

Note: An earlier version of this page presented numbers on prevalence, brain changes, comorbidity and treatment effects. We could not trace them to sources, so we removed them. What remains is labeled as draft or as sourced.

What this page is, and what it is not

Researchers and clinicians are still arguing about whether problem use of AI chatbots is a condition of its own, a form of an existing one, or a symptom of something else. Nothing here settles that. The page shows one way the question could be framed, using the language of the DSM, so that readers can see what a formal definition would have to cover.

Only one comparable behavioral condition is in the DSM-5-TR: internet gaming disorder, listed as a condition for further study. The World Health Organization’s ICD-11 includes gaming disorder, described as impaired control, growing priority over other interests and continued use despite harm, normally evident for at least 12 months. AI chatbots are in neither.

Diagnostic Criteria

A. Persistent Pattern of AI Use (draft)

The nine items below are adapted from the proposed internet gaming disorder criteria in the DSM-5-TR, with gaming replaced by AI chatbot use. We made that swap. Nobody has tested it. For gaming, the DSM proposes five or more of nine in a year, along with significant impairment or distress.

Preoccupation with AI interactions
Withdrawal symptoms such as sadness, anxiety or irritability when AI is unavailable
Tolerance: needing more time with AI to feel satisfied
Unsuccessful efforts to cut back or stop
Losing interest in other activities because of AI use
Continuing to use AI despite problems it causes
Hiding or lying about how much you use it
Using AI to escape or relieve negative moods, such as guilt or hopelessness
Having jeopardized or lost a job or relationship because of AI use

B. Duration and C. Functional Impairment

Any real definition would need a time period and clear harm to daily life. Hours of use alone do not define a problem. What matters is distress and harm.

D. Exclusion

Heavy chatbot use can be a symptom of, or a way of coping with, other problems such as depression, social anxiety or loneliness. A clinician has to sort that out. We can’t.

Specify Severity

An earlier version of this page graded severity by counting criteria, and described remission periods. We removed both. No study supports cut-offs for chatbot use, and counting boxes gives a false sense of precision.

Associated Features Supporting Diagnosis

Neurobiological Features

We found no brain imaging or chemical study of AI companion use that supports claims about specific brain regions or “dopamine pathways.” Earlier versions of this page made such claims. They are gone. The National Institute of Mental Health notes that the brain keeps developing into the mid-to-late 20s. That is a reason for care with young users, not evidence about chatbots.

Psychological Features

A four-week MIT Media Lab trial of nearly 1,000 adults found that long daily use was linked with worse outcomes, and that people with a stronger tendency toward attachment were more likely to see negative effects. The authors say these are correlations, not causal, and not peer reviewed.

Behavioral Patterns

A study accepted at CHI 2026 read 318 Reddit posts by users who said they were 13 to 17. The authors describe attachments with features of behavioral addiction, such as withdrawal, tolerance and relapse, plus lost sleep, falling grades and strained relationships. It shows what some teens say about themselves. It does not show how common this is.

Prevalence

We found no reliable estimate of how many people meet any definition of AI chatbot addiction. Surveys measure use, not dependence. Pew found that 64% of US teens use an AI chatbot and about 3 in 10 use one daily. Use is not addiction. Earlier versions of this page cited prevalence figures from other topics, such as internet addiction. They do not apply to AI chatbots, so we removed them.

Development and Course

Nothing reliable is known about how this develops or how long it lasts. We found no long-term studies. Claims you may see elsewhere about “stages” or timelines, including some we published before, are not backed by research.

Recovery Patterns

Some people cut back on their own. Others get support from friends, peer groups or a counselor. We found no published trials of a treatment designed for chatbot dependence, so we can’t say what works best or how long it takes.

Risk and Prognostic Factors

Temperamental Factors

The MIT trial above points to a tendency toward attachment as one risk. The Reddit study points to loneliness and wanting emotional support. These are starting points for research, not a list of causes.

Environmental Factors

Access may matter: many of these apps are free to start and always available. Pew’s 2025 survey shows how widely teens already use chatbots. Whether any of that causes harm is not settled.

Genetic and Physiological Factors

We found no research on genetic or physiological risk for chatbot dependence.

We found no cross-cultural research on problem chatbot use that we can responsibly summarize. Earlier claims about East Asian, Middle Eastern and Western differences were unsourced and have been removed. Attitudes to technology, family and mental health differ across cultures. Any formal definition would need to be tested in different groups before anyone applied it.

We found no reliable research on gender differences in problem chatbot use. Earlier claims about women and men were unsourced and have been removed.

Functional Consequences

People who describe a problem often mention lost sleep, falling behind at work or school, and pulling back from friends and family. The Reddit study above reports sleep loss, declining grades and strained relationships among teens who said they overused a companion app. These are reports from individuals. They show what people say, not what the technology does to the average user.

Differential Diagnosis

Many things look like chatbot overuse from the outside. Depression, social anxiety, ADHD, autism and loneliness can each lead to heavy use, and a strong interest in AI is not a disorder. This is why a professional assessment matters. Our own free assessment can’t tell these apart.

Comorbidity

We found no studies of how often chatbot overuse occurs alongside other conditions. Earlier percentage ranges on this page were unsourced and have been removed. In practice, a clinician looks at the person’s whole situation, not at the chatbot alone.

Treatment Implications

No treatment has been tested for AI chatbot dependence, as far as we found. A counselor will usually use methods they already use for compulsive behaviors and for any condition underneath, such as cognitive behavioral therapy. Earlier statistics about treatment effects were not drawn from AI chatbot studies and are removed. Our guide to finding counseling explains how to look for a licensed professional. The American Academy of Child and Adolescent Psychiatry says to talk to a doctor if you can’t moderate AI use or notice changes in mood.

Research Foundations and Verification

The sources on this page are the ones linked in the text: the American Psychiatric Association, the World Health Organization, the National Institute of Mental Health, the MIT Media Lab, Pew Research Center, the CHI 2026 Reddit study and the American Academy of Child and Adolescent Psychiatry. We read each of them in October 2026. We did not re-verify the other studies this page used to cite, so we have removed them.

This draft is an editorial exercise, not a clinical document. Do not cite it as a diagnostic standard. If you believe your AI use is harming your life, please see a qualified mental health professional.