Every therapist, social worker, counselor, and nurse remembers the first time theory has to become practice.
The concepts from class are there. The frameworks are there. A student may understand empathy, reflection, Motivational Interviewing, CBT, ethics, and therapeutic communication. But in the moment, sitting across from a client, knowledge alone is not always enough.
That is why more clinical education programs are turning to AI-simulated clients. The promise is clear: students can practice before entering high-stakes clinical settings, make mistakes in a low-risk environment, repeat difficult conversations, and build confidence before working with real clients.
But a simpler question comes first: does talking to an AI client actually make someone better? A recent Stanford-led study suggests the answer is — not on its own.
What the study found
Researchers ran a 75-minute randomized study with 94 novice counselors. One group practiced with an AI-simulated patient alone. The other practiced with the same AI-simulated patient and also received structured, AI-generated feedback.
The difference was meaningful. The practice-plus-feedback group improved in key client-centered microskills: their use of reflections increased by 3.6%, and their use of questions increased by 6.59%. Their empathy held steady and trended toward improvement.
The practice-only group showed a different pattern. Reflections and questions did not significantly improve, and empathy declined by 9.6%. The between-group difference in empathy change was 15% — also statistically significant.
The lesson is important for clinical education. AI can create a realistic practice conversation. But practice alone is not the same as learning. Without feedback, a learner may repeat the same habits, miss key opportunities for reflection, or drift toward problem-solving instead of client-centered listening.
“Talking to an AI client is the easy part. What matters is what happens after the conversation — whether the learner understands what they did well, what they missed, and how to respond differently next time.”
Practice is not enough
The study points to a distinction many educators already understand: repetition helps only when it is paired with reflection and guidance.
A student can ask more questions without asking better questions. They can speak more confidently without becoming more therapeutic. They can feel active in a conversation while still missing empathy, attunement, collaboration, or emotional reflection.
That is why feedback matters. Structured feedback turns a simulated session into a learning loop. It helps students notice patterns they may not see on their own, gives faculty and supervisors more concrete evidence to discuss, and connects practice back to the competencies a program already teaches.
In clinical education, the goal is not just to get students into more simulated conversations. It is to help them leave each conversation better prepared for the next one.
The same lesson, playing out at Frontier Nursing University
Faculty at Frontier Nursing University encountered a similar challenge in their psychiatric mental health nurse practitioner (PMHNP) program. In a distance-based program, students may not always have easy access to in-person standardized patients, role-play partners, or repeated low-stakes psychotherapy practice before entering clinical training.
Over a 30-day pilot, faculty tested TMind AI across cases involving Motivational Interviewing, Cognitive Behavioral Therapy, and general therapeutic communication. They examined whether AI-simulated psychotherapy sessions, paired with feedback, could realistically support skills training and formative assessment inside the curriculum.
The pilot was presented at the 2026 National Organization of Nurse Practitioner Faculties conference, and the faculty team is working toward a journal article, with plans to present again after collecting more student-use data.
The setting differs from the Stanford-led study. But the training principle is the same: simulation gets students into the room; feedback helps them improve.
Where TMind AI comes in
This is the idea TMind AI was built around. Students practice with realistic AI-simulated clients — but the platform doesn’t stop there. Every session comes with structured feedback tied to the competencies their program already teaches, so practice turns into real, trackable growth instead of just repetition.
It’s also where the name comes from. TMind is short for Theory of Mind — the human ability to understand that other people think, feel, and see things differently than we do. That’s the skill at the heart of empathy and clinical presence, and it’s what the platform is meant to help students build, not replace.
Early signals from real programs
TMind AI is currently being used, piloted, or explored across a growing group of universities, health systems, residency programs, and behavioral health organizations:
Salisbury University
Simmons University
University of South Carolina
Delaware State University
Texas Christian University
University of Houston
University of Texas at San Antonio
University of Texas at Tyler
Baptist Health / University of Arkansas for Medical Sciences
Creighton University
Connecticut Institute for Communities
Medical College of Wisconsin
Frontier Nursing University
The platform is also shaped by an advisor group of deans, professors, and clinicians from Rutgers School of Social Work, Simmons University, Salisbury University, Stanford’s Department of Psychiatry and Behavioral Sciences, and Lindsey Wilson University.
At Simmons University School of Social Work, faculty piloted TMind AI in SWO 421A: Generalist Practice, a pre-practicum course, to explore whether simulated-client practice could strengthen MSW student confidence, practicum readiness, and applied clinical skills. Grounded in Social Cognitive Theory, the pilot evaluated student performance across empathy, cultural humility, client-centered practice, and ethics aligned with the NASW Code of Ethics.
Instructor focus groups reported that students moved from theory into applied practice, gained confidence, showed reduced imposter syndrome, and demonstrated stronger reflection. The early findings were qualitative and directional rather than statistically confirmed, and the study has since been expanded under IRB approval across additional cohorts.
“The hands-on learning with the simulation client really helped to embed an understanding of foundational clinical skills in a much more authentic way.”
- Participating instructor, Simmons University
“The students had the opportunity to apply generalist practice skills in a way that demonstrated collaboration with the client. Beyond their book learning, they were directly applying it clinically through simulation.”
- Participating instructor, Simmons University
These observations align with the broader research direction: AI simulation is most valuable when it helps learners move from passive knowledge into active, reflective, client-centered practice.
The point
Whether an AI can hold a conversation was never really the question. Whether it can help someone walk away better than they walked in — that’s what matters. And that’s what TMind AI is building, alongside the programs testing it in real classrooms.