Self-check guide / By the editorial team
Information only, not medical advice or a diagnosis. In a crisis in the US, call or text 988.
You open Claude for one quick task. Three hours later you have asked it about a work report, a text to a friend and what to cook for dinner.
If that sounds familiar, you are not unusual, and you are not necessarily in trouble. Plenty of people use Claude heavily for work and feel fine about it. The question this page helps with is narrower: does your use still feel like your choice?
Two things to know about Claude. It is an AI model made by Anthropic, not a person. And Anthropic’s terms say you must be at least 18, or the higher age of consent where you live (Anthropic).
One more honest point: “Claude addiction” is not a diagnosis. The APA’s DSM-5-TR includes gambling disorder as its sole behavioral addiction and flags internet gaming disorder for further study (APA). The phrases “productivity AI dependency” and “cognitive outsourcing,” which earlier versions of this page attributed to researchers, are not established terms, so we no longer use them as if they were. A habit can cost you sleep, time or confidence without a label attached.
In crisis or thinking about hurting yourself? US readers can call or text 988 (988 Lifeline, always open) or text HOME to 741741 (Crisis Text Line). In immediate danger, call your local emergency number.
Signs You Might Be Addicted to Claude AI
No single sign proves anything. Look for a pattern that lasts, and ask whether it costs you something you care about. Three plain questions do most of the work:
- Do you use Claude more than you plan to?
- Have you tried to cut back and could not?
- Is it affecting your sleep, job, studies or relationships?
Behavior
- Checking on decisions you used to make alone. Small ones first, then bigger.
- Nerves before you start. A task feels risky until you have asked Claude about it.
- Quality worry. You fret that anything unpolished by Claude is not good enough.
- Postponing choices until you can get its view.
- Work stalls when Claude is down or you are offline.
Confidence and Thinking
- Fewer ideas of your own. Brainstorming alone feels flat.
- Doubting your judgment in areas where you have real experience.
- Relying on it to recall things you used to remember.
- Hard problems feel heavier than they did. Be careful here. Stress, poor sleep and distraction can do the same, so this sign alone says little.
Work and Social Life
- Plateau. You feel your skills are not growing because Claude does the stretch tasks.
- Feeling like a fraud. You wonder what part of your output is really yours.
- Fear of unassisted work, like a live meeting or a handwritten exam.
- Personal use. You use it for messages to people close to you or when you feel lonely.
See yourself in a few of these? The free assessment is a sensible next step. It gives you a self-check and no diagnosis.
The 5 Types of Claude AI Dependency
These work-use labels are ours, invented to help you describe a habit. Nothing clinical or researched stands behind them.
1. The Strategic Delegator
You use Claude to plan, weigh options and think through business problems. You feel unsure about strategic calls without it.
2. The Communication Perfectionist
Emails, reports and slides all go through it. Your own wording feels inadequate next to its polish.
3. The Analysis Dependent
You hand it data interpretation and research synthesis, and trust your own reading of the numbers less.
4. The Creative Collaborator
Brainstorming and content ideas start with Claude. Starting without it feels blocked.
5. The Process Optimizer
You build Claude into your workflows and task systems, and feel disorganized without it.
Why Claude AI Use Can Become Hard to Cut Back at Work
We can’t tell you what caused your habit. These are common reasons people give, and none is proven:
- Pressure. If colleagues use AI and output expectations rise, using it can feel required.
- Better first drafts. When its draft beats yours, you stop writing your own.
- Quick relief. It removes the discomfort of uncertainty right away.
- Always there. It does not get tired or busy.
Research on heavy chatbot use is thin. A four-week trial with 981 adults found no significant effect from the settings researchers assigned, yet people who chose heavier use reported more loneliness and emotional dependence (arXiv). It showed a link, not a cause, and it tested ChatGPT. We have found no study of Claude dependency.
The Hidden Costs of Claude Dependency
What follows are possible costs to check in yourself. They are not predictions.
Thinking and Learning
- Less practice at analysis, writing and decisions you now hand off
- Less learning when you skip the struggle with new material
- Less chance to notice errors if you accept answers unchecked. Models can be wrong, so check what matters.
Work
- Unclear rules. Many workplaces and schools set rules on AI use or require disclosure. Read yours.
- Sensitive data. Check what your employer allows you to paste into an AI tool.
- A dependence on one tool you do not control.
Personal
- Trusting yourself less
- Less time with friends and colleagues
- Less tolerance for waiting and not knowing
Most of this can change. We have no data on how fast, so don’t plan around a timeline.
Breaking Free: Professional Recovery Strategies
This plan is a suggestion. We picked the week ranges for convenience, and no clinician set them. Many people want to use Claude differently, not stop.
A Gradual Plan for Work Use
Weeks 1-2: Notice. Track what you use Claude for and how you feel just before. Add a 15-minute “try it myself” delay for tasks you know well.
Weeks 3-4: Draft first. Write the full email or outline yourself, then ask for feedback. Make a decision, then compare with Claude’s view.
Weeks 5-6: Guard a block. Keep one AI-free work period for your hardest task. Present one piece of unassisted work and notice how it goes.
Exercises by Skill
Strategy. Write your plan on one page before consulting anything. Keep a note of your calls and how they turned out.
Communication. Send routine emails unpolished. Ask a colleague for feedback on bigger pieces.
Analysis. Look at the raw numbers or source first. Write your own two-line takeaway before you ask for one.
Use Claude’s Settings
Claude’s memory is on by default for Free, Pro and Max accounts. You can pause it, reset it, or start an incognito chat that is not saved (Claude Help Center). Anthropic says its models still need to improve at not simply agreeing with users (Anthropic), so don’t treat Claude’s approval as proof you are right.
What the Assessment Tells You
The assessment has 20 questions and works as a self-check. No one has validated it as a clinical instrument, and it diagnoses nothing. Adults only. The result is emailed to you.
Your Band
Your total score lands in Low, Caution or High. Treat the band as a pointer. A Low result does not rule out a problem.
What Your Result Means
The email explains your band in plain language.
Next Steps
You get small steps to start today and a note on when to talk to a doctor or counselor.
What the Assessment Cannot Tell You
Earlier versions of this page quoted people about their results. We could not confirm those people agreed, so we removed the quotes and we do not publish personal stories without confirmed consent.
The assessment also cannot say whether your work use is excessive. That depends on your job, your rules and your goals. If you are unsure, talk to someone you trust or a counselor.
Illustrative stories
These are illustrative composites written by our team to show common experiences. They are not quotes from named readers and do not describe results you should expect.
Thinking first
“I use Claude to plan my work, and it’s good at that. I noticed I wasn’t thinking problems through first anymore. I started by spending a few minutes with pen and paper before I asked, and then used Claude to test my own plan.”
Telling it everything
“I like how thoughtful the replies feel, and I’ve caught myself telling it things I hadn’t told people. I’m not ashamed, but I wanted some balance. I started by talking one worry through with a friend before typing it up.”
Frequently Asked Questions
Click each question to expand the answer
Is Claude dependency different from ChatGPT dependency?
We have no research comparing them. The habits people describe look similar. Anthropic says only a small share of Claude.ai conversations involve emotional support or companionship, 2.9% of the free and Pro conversations it studied, and that its study could not assess emotional dependence (Anthropic).
How long does it take to rebuild skills?
No research gives a timeline. Skills return at different speeds. A counselor can help if you stall.
Can I still use Claude at work?
Yes. A common pattern is a final read-through from Claude after you write the first draft yourself. Check your employer’s AI policy first.
Are some jobs riskier?
We don’t have data that says so. Jobs heavy on writing and analysis give more chances to use it, which is a different thing from higher risk.
What is the difference between healthy use and a problem?
Comfortable use leaves you able to work without it. Trouble looks like going past your plan, failing to cut back, or losing sleep, work time or closeness with people. A diagnosis is not required to take that seriously.
What happens to my assessment answers?
The result is sent to the email you enter, and you can opt in to follow-up emails and stop them whenever you like. See the privacy policy first.
Medical Disclaimer
General information only, not medical advice, diagnosis or treatment. If you feel severe anxiety without AI, your work is slipping, or you think about self-harm, seek professional help promptly.
Crisis resources: 988 Suicide & Crisis Lifeline: call or text 988 | Crisis Text Line: text HOME to 741741 | Find a therapist: psychologytoday.com