Ethical AI Use for Social Work Students

AI is coming at you from everywhere, and honestly? Some of it is legitimately useful for your clinical development. Some of it is a confidentiality nightmare. The trick is knowing the difference.

I'm not going to tell you AI is evil or that you shouldn't use it. That's not realistic. But I will tell you exactly what you can and can't do without compromising your clients, your career, or your ethics.

What AI Can Actually Help With

Case conceptualization practice. You can write a fake case (with nothing real about it) and ask AI to help you think through different theories. "What would a systems theorist notice about this family dynamic?" That's great. It pushes your thinking.

Learning clinical terms and concepts. If you're confused about the difference between dissociation and depersonalization, AI can explain it. If you need to understand psychopharmacology basics before talking to a psychiatrist, AI is a quick resource.

Assignment planning. You can brainstorm paper structures, outline ideas, plan your time. You're still writing the paper. You're just using AI as a thinking partner.

Skill drills. You can practice writing progress notes, case notes, treatment plans using fake information. You can ask AI to critique your note and explain why. This is studying.

Theory exploration. You want to understand attachment theory better? Ask AI to explain it. Read it. Read your textbook. Compare what you learned.

Reducing repetitive busywork. Your professor asks you to list every state's HIPAA compliance requirements. AI can get you started so you're not manually googling each one. You still verify the information and understand it.

What You Absolutely Cannot Do

Do not enter any real client information. Not even anonymized versions. Not "client with substance use disorder and ADHD and family conflict." You don't know what patterns AI might pick up on that could theoretically re-identify someone. This isn't paranoia. It's the standard.

Do not describe your supervisor's feedback about specific real cases. "My supervisor said I was too harsh with my client about their drinking. Here's what happened:" is too much. That's identifying information about the supervisor, the client, and the setting.

Do not ask for actual diagnosis advice or clinical decision-making. You're in practicum for a reason. You're not the decision-maker yet. Your supervisor is. Use AI to understand diagnoses. Don't use it to decide what your client should do.

Do not submit AI-written work as your own. Your professors will catch it. More importantly, you're paying for an education. Submitting ChatGPT output as your paper defeats the purpose.

Do not use AI to replace supervision. You're confused about something in session? That's for your supervisor. Not AI. AI doesn't know your client, your agency, your supervisor's expectations, or what actually happened.

Do not share recordings, session transcripts, or detailed case notes. Ever. Not even "just this part." Not even "with identifying information removed."

How to Remove Identifying Details (If You Really Need To)

Sometimes you need to run something by AI — like you're stuck conceptualizing and you want another angle. Fine. But strip it:

• Remove names, ages, specific locations
• Remove dates and timeline details that could identify them
• Remove specific diagnoses if they're rare
• Remove any details about your agency or your supervisor
• Remove anything about family members or other identifying relationships
• Change genders or create fictional family structures if it doesn't change the conceptual question

Notice: the second version asks the actual question you have without any identifying information.

Why AI Output Still Needs Human Judgment

AI is good at patterns. It's bad at context. It doesn't know your client. It doesn't know the nuance of your relationship. It doesn't know what's actually true in someone's life versus what's statistically common.

If AI suggests something that doesn't match what your supervisor said, what your textbook said, or what you know about your client's actual life — trust your supervisor and the evidence. Not the AI.

AI is a tool for thinking, not for knowing.

Academic Integrity Concerns

Check your school's AI policy. Some schools are specific about where AI is allowed and where it isn't. Some are still figuring it out. But the general rule: if you use AI, disclose it. Some professors want you to show the prompt you used. Some want you to explain how you used it. Some don't allow it in certain assignments.

When in doubt, ask your professor or advisor. They'd rather clarify than catch you later.

The Real Talk

Using AI doesn't make you unethical. Hiding that you used it does. Entering client information does. Letting AI make clinical decisions does.

You're in grad school to develop your own clinical thinking. AI can support that. It shouldn't replace it. And it never, ever touches your clients' information.

Related: Motivational Interviewing for Social Work Students · First Practicum Session Checklist