Clinical AI Dependency Assessment Scale (CAIDAS)

A plain description of how the CAIDAS works and what it can and cannot tell you.

RESEARCH INSTRUMENT DISCLAIMER

CAIDAS is a self-reflection screening tool written by The AI Addiction Center. It is not a diagnostic test, and it has not been through psychometric validation or clinical testing. Do not use it to diagnose yourself or anyone else, or to decide on treatment. A licensed professional has to make those calls.

Download the CAIDAS PDF

A PDF is available from the link below. It is a separate document from the online assessment. Its item count, wording and scoring can differ from what this page describes, so read the PDF’s own cover page before you use it. Check the download page for its terms.

→ Download CAIDAS PDF

Executive Summary

CAIDAS is the scale behind the free AI addiction assessment. The online version has 20 questions about time, habits, mood, sleep and relationships. Each question has five answer options. The scores are added up, with no weighting, to a total from 0 to 79. The total falls into one of three bands: 0 to 26 Low, 27 to 52 Caution, 53 to 79 High.

The scoring was updated in July 2026. We do not publish the earlier version. Treat that as a limit: you can’t compare the two.

The bands sort your answers. They do not diagnose anything. High means many answers were at the “often” end. Low does not rule out a problem.

Theoretical Foundation

Researchers who study behavioral addictions often describe them with a few shared features. The best-known list is the components model of addiction (Griffiths, 2005): salience, mood modification, tolerance, withdrawal, conflict and relapse. Other work looks at why some people attach to non-human things. Epley and colleagues (2007) describe a general human tendency to treat things as if they have human minds.

We mention these because they are the ideas behind how many people think about problem use. We have not published an item-by-item mapping from CAIDAS questions to any of them, and this page does not claim one. It also does not claim CAIDAS matches DSM-5-TR criteria. AI chatbot addiction is not a formal diagnosis. The DSM-5-TR lists only internet gaming disorder, as a condition for further study, and ICD-11 has gaming disorder among its recognized conditions. Neither covers AI chatbots.

Assessment Structure

Scale Specifications

• Items: 20 questions
• Response format: five answer options per question
• Scoring: points are added up as a flat sum, total range 0 to 79
• Bands: 0 to 26 Low, 27 to 52 Caution, 53 to 79 High
• Language: English
• Audience: adults, 18 and over

We have not timed the assessment, so we don’t state how long it takes.

What the questions cover

The questions ask about time spent, difficulty cutting back, mood, sleep, and effects on relationships and responsibilities. The assessment covers AI use in general, so you answer for all your apps together.

Scoring and Interpretation

Score bands

The cut-off points were set by our editors and reviewed in July 2026. They have not been tested against any outside measure of problem use.

0 to 26, Low: most of your answers were at the “rarely” end. That does not rule out a problem.

27 to 52, Caution: some answers point to a habit worth watching.

53 to 79, High: many answers were at the “often” end. Consider talking to a doctor or counselor.

None of these is a verdict. If the band and your own sense of things disagree, talk to someone.

Psychometric Development Standards

A scale like this is only trustworthy once it has been tested. A full validation study normally checks, among other things:

• Reliability: do people get similar scores when they answer twice, and do the questions hang together?
• Factor structure: do the questions group into the parts they are meant to measure?
• Validity against other measures: do scores line up with established scales and with outcomes such as sleep or functioning?
• Cut-off accuracy: are the bands placed where they separate people with real problems from people without?

DeVellis’s textbook on scale development (2017) describes these steps. CAIDAS has not completed them. We have no published figures for any of these checks, so we do not state any.

Intended Use

Who it is for

CAIDAS is meant for self-reflection and for starting a conversation, for example with a doctor or counselor. It is not meant for screening patients in a clinic, planning treatment, tracking progress, workplace evaluation or research on how common AI dependence is. If a researcher wants to use it, they should contact us first and validate it themselves.

Where it fits

Established tools exist for internet and social media use, such as Young’s Internet Addiction Test (1998) and the Bergen Social Media Addiction Scale. Those have been studied far more than CAIDAS. If you need a measure for research or clinical use, choose one with published validation.

Current Research Status

CAIDAS has not had a published validation study, and we have no data on reliability or accuracy. We have the completions from our own site, which come from people who chose to take it and are not a sample of the population.

If that changes, we will say so here and give the details.

Methodological Approach

The questions are written by The AI Addiction Center. This page does not list them. We changed the scoring after the July 2026 review. Earlier versions of this page described a longer 48-item version with six scored dimensions and a panel of 12 specialists. We could not document those details, so we removed them. The scale may be revised again, and we will update this page when it is.

Comparison with Existing Instruments

Most published scales for technology use were built for gaming, social media or the internet in general. AI chatbots differ in some ways: they answer you, remember you and are always available. That is why we wrote a scale for them. It is also why we can’t lean on the older tools’ validation. Their results do not transfer to CAIDAS.

Administration Guidelines

• Take it somewhere private and answer honestly. There are no right answers.
• Answer about your overall AI use, across all the apps you use together.
• Adults only. If you are under 18, talk to a parent, carer or another adult you trust.
• Don’t use it during a crisis. If you are in crisis, call or text 988, or text HOME to 741741 (Crisis Text Line).
• If you score High or the result worries you, talk to a doctor or counselor. Our guide to finding counseling explains where to look.

Ethical Considerations

• Purpose and limits are stated up front: screening for reflection, not diagnosis.
• Your result is sent by email, not shown on screen. How we handle that email is in the privacy policy.
• We try not to treat normal technology use as a problem. Plenty of heavy users are fine.
• Usage patterns differ across cultures and ages. A single score can’t capture that.
• The assessment is for adults only, because it has not been tested with young people.

Citation and Usage

To refer to CAIDAS, cite it as: The AI Addiction Center. Clinical AI Dependency Assessment Scale (CAIDAS). Web page, reviewed October 2026. Please state that it is an unvalidated screening tool. To ask about other uses, contact us through the contact page.

Download the CAIDAS PDF

→ Download CAIDAS PDF

Read the PDF’s cover page first. It may describe a different version from the online assessment.

References

1. American Psychiatric Association. (2022). Diagnostic and Statistical Manual of Mental Disorders (5th ed., text rev.).

2. Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864-886.

3. Griffiths, M. (2005). A “components” model of addiction within a biopsychosocial framework. Journal of Substance Use, 10(4), 191-197.

4. DeVellis, R. F. (2017). Scale Development: Theory and Applications (4th ed.). SAGE Publications.

5. International Test Commission. (2017). The ITC Guidelines for Translating and Adapting Tests (2nd ed.).

6. World Health Organization. Gaming disorder (ICD-11, 6C51).

7. Young, K. S. (1998). Internet addiction: The emergence of a new clinical disorder. CyberPsychology & Behavior, 1(3), 237-244.