Skip to content

Evidence · Australian counsellor training

The evidence for practising with simulated clients.

Counselling skills grow through practice: many repetitions, with feedback, before the stakes are real. Role-plays with classmates are valuable but hard to scale. This page sets out what published research shows about AI simulated clients in counsellor training, where they fit in Australian training, and what the evidence does not show yet.

Scope: simulated clients that counselling students practise with during training. Nothing here is about AI acting as anyone’s counsellor.

What the research shows

Six findings, in plain words.

Five recent studies of AI simulated clients in counselling and psychology training, and one study of how effective therapists keep improving.

  1. 41 trainees · 3 sessions

    Confidence rose, and stayed up.

    Self-efficacy rose across three practice sessions with an AI client and held after training. Anxiety fell, mostly after the first session.

    (Jeong et al., 2025)Paper

  2. 112 students · 8 weeks

    Skills grew with practice.

    Educational psychology students who practised with a simulated client gained therapeutic skills, compared with a group who did not.

    (Kakabayeva et al., 2025)Paper

  3. Qualitative study

    Low pressure, at their own pace.

    Trainees said practising with an AI client eased their nerves before practicum. They also named its limits, such as a client who agreed too readily.

    (Akkurt et al., 2025)Paper

  4. 24 psychology trainees · voice vs text

    A voice felt more useful than text.

    Trainees who tried both rated a speaking avatar client above a text chatbot for usefulness and for how much their skills improved. Self-rated, and not yet peer reviewed.

    (El Sawah et al., 2025, preprint)Paper

  5. 16 clinicians · 29 sessions

    Believable, but too easy.

    Clinicians found the AI clients’ stories convincing, but most felt the clients seemed robotic, and some called them too compliant. Realistic resistance is the hard part.

    (Cabrera Lozoya et al., 2025)Paper

  6. 2.8× the solitary practice

    Practice, not years.

    Among practising therapists, time in solitary deliberate practice tracked client outcomes; years of experience did not. Self-report, and not students: a direction, not a dose.

    (Chow et al., 2015)Paper

Built for Australian training

Made in Adelaide. Mapped to PACFA and the ACA.

Made in Adelaide

As far as we know, CounselSim is the first AI client tool built in Australia and mapped to PACFA and ACA standards.

PACFA Training Standards (2022)
“preparation for client work within the training context”
That is how PACFA places role-plays and practice sessions, separate from the supervised work with real clients that placement requires (PACFA, 2022). CounselSim sits in the same place: a practice session never counts towards placement, and it does not replace supervision.
Two standards, one report

Feedback is mapped to both Australian frameworks, so programmes under either body read the same report.

PACFA competencies (2026)ACA Code of Ethics and Practice

Practice clients live across Australia and come from many cultures, all adults, aged 18 to 77.

Where it stands

An early field, moving quickly.

We would rather you weighed the evidence as it is.

The field is moving
  • Touro University’s School of Health Sciences in New York began training counselling students with AI simulated clients in April 2026, in its course “Counseling the Whole Person” (Touro University, 2026).
  • In Australia, the Psychological Society described virtual client simulations as a complementary tool in training, not a replacement for working with people (APS, 2026).
  • The national standards for counsellors and psychotherapists, published in October 2025, set supervised placement with real clients and say nothing about simulation (Department of Health, 2025). CounselSim prepares students for that placement; it does not stand in for it.
CounselSim: built in Australia
What the research doesn’t show yet
  • 1

    Samples are small, and most studies come from a single programme.

  • 2

    Most outcomes are self-reported, such as confidence and anxiety, with short follow-up.

  • 3

    No study yet links practice with an AI client to outcomes for real clients.

  • 4

    The studies set up general chatbots as clients. Their findings may not carry over to every tool, including ours.

  • 5

    AI clients do not stand in for supervised placement, or for practising with people.

In CounselSim

Built around deliberate practice.

Students practise whole sessions out loud with 27 adult clients, each with their own history. Four parts of the design follow from the research.

Repetition, one skill at a time

Return to the same client and the same moment as often as needed, so practice is deliberate rather than occasional.

Replay, with notes on the moment

Every practice session is recorded. Students pin notes to the exact second, and supervisors reply there.

Feedback mapped to the standards

Feedback on ten counselling skills, mapped to PACFA competencies and the ACA Code. Every AI score is marked suggested.

A supervisor decides

Supervisors review the recording and give their own rating. The AI’s suggestion starts that conversation; it never ends it.

Evaluate it with us.

We welcome Australian universities who want to evaluate CounselSim with their own students, on their own terms. It starts with a 30-minute call.

Book a demo

References

Checked 27 October 2026.

  1. Australian Psychological Society. (2026, May 6). How virtual client simulations are aiding psychology training. https://psychology.org.au/insights/virtual-client-simulations-aid-psychology-training
  2. Cabrera Lozoya, D., Conway, M., De Duro, E. S., & D’Alfonso, S. (2025). Leveraging large language models for simulated psychotherapy client interactions: Development and usability study of Client101. JMIR Medical Education, 11, Article e68056. https://doi.org/10.2196/68056
  3. Department of Health, Disability and Aged Care. (2025, October 13). National standards for counsellors and psychotherapists: Summary report. Australian Government. https://www.health.gov.au/sites/default/files/2025-10/national-standards-for-counsellors-and-psychotherapists-summary-report_1.pdf
  4. El Sawah, V., Bhardwaj, A., Pryke-Hobbes, A., Gamaleldin, D., Ang, C. S., & Martin, A. K. (2025). Artificial intelligence as a training tool in clinical psychology: A comparison of text-based and avatar simulations [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2601.11533
  5. Akkurt, M., Maurya, R. K., & Brown, C. (2025). Learning through simulation: Counselor trainees’ interactions with ChatGPT as a client. Behavioral Sciences, 15(12), 1660. https://doi.org/10.3390/bs15121660
  6. Chow, D. L., Miller, S. D., Seidel, J. A., Kane, R. T., Thornton, J. A., & Andrews, W. P. (2015). The role of deliberate practice in the development of highly effective psychotherapists. Psychotherapy, 52(3), 337–345. https://doi.org/10.1037/pst0000015
  7. Jeong, D., Lee, H., Yang, H., Park, J. C., & Lee, S. M. (2025). AI integration in counseling training: Aiding counselors-in-training in self-efficacy enhancement and anxiety reduction. Journal of Counseling & Development, 103(3), 346–360. https://doi.org/10.1002/jcad.12560
  8. Kakabayeva, D., Abibulayeva, A., Orazbayeva, K., & Naviy, L. (2025). AI as a simulated client: Impact on educational psychology students’ therapeutic skills. Journal of Curriculum Studies Research, 7(2), 31–52. https://doi.org/10.46303/jcsr.2025.10
  9. Psychotherapy and Counselling Federation of Australia. (2022). PACFA professional training standards. https://pacfa.org.au/common/Uploaded%20files/PCFA/Documents/Training%20Standards/PACFA_TrainingStandards_2022.pdf
  10. Touro University School of Health Sciences. (2026, April 17). Future counselors train with AI avatars at Touro’s School of Health Sciences. https://shs.touro.edu/news/stories/future-counselors-train-with-ai.php