Stuck in a Therapy Role-Play? How to Use In-Session AI Coaching Without Losing Your Voice

A simulated client says, “There's no point talking. No one ever understands me.” You understand empathy in theory, but no response comes to mind. Ending the exercise wastes a learning opportunity. Copying an AI-generated line, however, may keep the conversation moving without giving you much practice.

The more useful question is not whether support is allowed. It is how to use support without outsourcing your clinical reasoning.

What is in-session coaching?

In a simulated encounter, coaching provides prompts or feedback while you practise responding. It does not tell you what you must say to an actual client. Suggestions are alternatives for educational consideration, not treatment decisions.

According to the current TheraMentor feature descriptions, three tools support this process:

• Response suggestion (the Greek interface calls this Ιδέα διατύπωσης): three optional ways to continue when you feel stuck.
• Message feedback (Δες σχόλιο): brief feedback tied to a response you have sent.
• In-session reflection (Στιγμιαίος Αναστοχασμός): a broader snapshot of the exchange and areas worth revisiting.

They serve different purposes: generating options, reviewing a specific intervention and stepping back to examine your overall approach.

Make your own attempt first

Return to “No one understands me.” Before opening a suggested response, try writing your own line:

“It sounds as though you've been feeling very alone in this.”

Now compare it with the alternatives. Instead of asking which sounds most polished, ask: Which response stays closest to the client's words? Which introduces an assumption? Which leaves room for the client to disagree or clarify?

In this approach, AI suggestions become comparison material, not a script to memorize.

Using message-level feedback thoughtfully

Feedback on an individual response is most helpful when you already have a learning objective. Suppose your focus is reflecting emotion. Review the suggestion against three questions:

1. Did I acknowledge emotion, or only ask for more information?
2. Did I present an interpretation as fact?
3. How could I express the same intention with fewer assumptions?

An AI comment is not an independent clinical competency assessment. Even plausible suggestions can miss context, timing or the client's meaning.

When to pause for in-session reflection

After several exchanges, you may notice that the conversation has become repetitive. Use in-session reflection as a reset: “What pattern am I repeating?” and “What one change will I try in the next three turns?”

If you have asked three questions in a row, for example, you might test a reflection followed by a little space. The educational value comes from deliberate comparison and revision, not automatic agreement with the tool.

A ten-minute practice routine

Before the simulation: pick one focus, such as non-judgmental exploration.

During: respond in your own words first. Open a suggestion only when needed, and check feedback after a challenging turn.

After: record one response you would keep, one you would revise and one goal for your next exercise.

Emerging research is worth watching without exaggerating it. A small 2026 study in a graduate psychotherapy course tested different combinations of AI feedback and reflection and found educational value for a particular design. It did not establish that every AI coaching tool improves real-world clinical performance.

Try it with TheraMentor

Start a simulated session, select a skill to practise and use response suggestions, message-level feedback and in-session reflection as learning aids rather than an autopilot. TheraMentor is designed for education, not as a substitute for professional supervision or clinical experience.

Next step: Practise with TheraMentor.