ChatGPT became public in late November 2022. Within a couple of years, counseling students were practicing with it, faculty were writing about it, and programs were trying to decide what to allow. I kept wondering what all of that writing actually added up to.
So Dr. Abigail Van Patten-Freeman and I set out to map it. Our scoping review, now in press at The Interdisciplinary Journal for Research on Pedagogy and Learning Design, looks at every peer-reviewed source on generative AI in counselor education we could find from January 2015 through March 2026. We searched four databases, hand-searched counseling journals, and charted each source in duplicate.
1,286
records screened from databases, hand-searching, and citation tracking
14
sources met our criteria: 3 conceptual and 11 empirical
0
published before 2024, even though our search window began in 2015
What we found
The simulated client dominates
Eight of the 14 sources used AI to play a counseling client. The AI held its role well, but three limits came up again and again: no nonverbal information, clients that were too idealized, and a culturally neutral default.
Culture has to be prompted
ChatGPT did not add racial or ethnic background on its own. Cultural content showed up only when someone wrote it into the prompt, so the cultural learning depends on whoever writes that prompt.
Two domains, lots of gaps
Most sources mapped to Counseling Practice and Relationships or to Assessment. Career Development and Group Counseling did not appear at all.
Policy is lagging too. One study we reviewed looked at 92 CACREP-accredited doctoral programs: 72 had some kind of AI policy, but only five had one written for counselor education.
The field has established what generative AI does and what trainees think, but nothing about whether trainees who use it become better counselors.
From our review
The question nobody has asked yet
Eleven of the sources were empirical, and that sounds encouraging until you look at what they measured. None tracked change in a trainee’s counseling competency. None used a comparison group, and none followed students over time. We know AI can produce a convincing client. We do not yet know whether practicing with one makes someone a better counselor.
Counselor education is well placed to find out. We already have validated skills measures, supervisors who rate recorded sessions, and a practicum sequence where learning either transfers or doesn’t. To help, we propose the Counselor Education AI Integration and Validation framework and a research agenda built around competency-based, comparative, and longitudinal studies. No source has yet tested AI in a crisis or risk-assessment scenario, which our review identifies as the most safety-critical gap.
If you teach counseling and you are using AI with students, I would love to hear what you are seeing. The next wave of this research needs to come from classrooms and practicum sites, not just from the tools.
The article
Van Patten-Freeman, A., & Karkhanis, S. (in press). Generative artificial intelligence in counselor education: A scoping review of applications, ethics, and CACREP competency mapping (2015–2026). The Interdisciplinary Journal for Research on Pedagogy and Learning Design.
I’ll add a link here once the article is published.
