AI Mentor vs Human Supervision in Therapy Training

Not all feedback is the same. Identifying that a trainee asked five questions in a row is one task. Exploring why the trainee felt uncomfortable with silence is another. Technology may assist with the first. The second often needs a thoughtful human supervisory relationship.

That is why “AI or supervisor?” is the wrong question. The important issue is how their responsibilities differ.

What can an AI mentor contribute?

In an educational simulation, AI may help learners select a narrow goal, revisit a specific response, or explore alternative wording. Because it can be available between classes or supervision appointments, it may support more frequent rehearsal.

Availability does not make it an expert witness to the learner's development. A system cannot reliably understand the whole personal and professional context, nor should it certify clinical readiness on its own.

What does a human supervisor bring?

A supervisor can examine the therapeutic relationship, ethics, the trainee's emotional responses, treatment context, and how skills develop over time. They can challenge assumptions, notice patterns across encounters, and take responsibility for educational judgement and professional standards.

Supervision also involves accountability, boundaries, and safety. Those responsibilities do not disappear because an automated tool can produce convincing commentary.

One simulated moment, two perspectives

During a training conversation, the fictional client says: “I don't want to talk about this anymore.”

AI feedback might observe that the trainee immediately asked three more questions. A supervisor could explore a different layer: “What did you feel when the client withdrew? What were you hoping those questions would achieve?”

These responses are not interchangeable. One marks an observable behaviour; the other relates it to the learner's clinical thinking and emotional experience.

What does recent discussion suggest?

A 2026 article from the American Psychological Association described the promise of more accessible practice and immediate feedback while warning that AI-generated comments may sound authoritative even when clinically questionable. A 2026 academic paper on AI in CBT training and supervision similarly examined opportunities alongside implementation and evaluation challenges.

The sensible model is complementarity with explicit boundaries, not automatic substitution.

A practical workflow for training programmes

A trainee completes a fictional simulation, selects a challenging exchange, reviews an AI suggestion as a hypothesis, and brings that moment to supervision. The supervisor remains accountable for how the exercise is interpreted and what educational decisions follow.

Institutions may also reasonably choose to use simulations with AI feedback disabled, depending on curriculum design and policy. Flexible educational settings should respect that distinction.

TheraMentor separates automated guidance from human supervisor feedback because each serves a different educational purpose. AI may expand practice opportunities; human responsibility remains essential.