The aim of today’s article is to share the introductory section of a study published in 2026 in the Journal of Psychology and AI, which proposes a theoretical framework for integrating artificial intelligence into the training, supervision, and delivery of Single-Session Therapy (SST).
Why present this study?
It is a common experience among professionals, both in clinical and educational settings, to encounter the growing presence of artificial intelligence without yet having clear guidelines regarding its limitations, potentialities, and the criteria required for its ethical and effective use.
Clients themselves increasingly bring this topic into the therapy room, reporting that they have used AI tools to seek guidance, search for answers, or process thoughts between sessions. However, their experiences reveal perspectives that are not always consistent. What initially appears to be advantageous, such as unlimited accessibility, clarity of responses, the absence of judgment, and the reduction of stigma associated with seeking help, does not always prove beneficial within the therapeutic process.
Unlimited accessibility, for example, may become a dysfunctional solution that prevents individuals from accessing their own internal resources for processing emotional experiences. Similarly, the perceived absence of judgment may, in some situations, limit the reflective function that emerges through engagement with another human being capable of offering a different perspective and broadening one’s understanding.
These are all aspects that deserve careful consideration, making it essential for professionals to develop a critical and informed perspective on such tools rather than adopting uncritical positions.
An Introduction to the SST-AI Framework (Joseph et al., 2026): What Is the Starting Point?
The authors Jasmine Joseph, Manesh Pai, Keigo Asai, and Magdalena Turek developed and published in March 2026, in the Journal of Psychology and AI, a theoretical proposal for integrating artificial intelligence into Single-Session Therapy while maintaining the human therapist as the central therapeutic agent (Joseph et al., 2026). Their work is grounded in two key questions:
“What would happen if, during an SST session, an artificial intelligence system silently listened to the therapeutic dialogue and discreetly suggested to the clinician, in real time, the precise moment in which the client was opening a window for change?”
“And what if, after the session, that same system automatically generated a supervision report, identifying the therapist’s strengths and missed opportunities?”
The proposed work is not an empirical study, but rather a theoretical and design-oriented contribution addressing the complex and highly relevant intersection between SST and AI. It represents one of the most ambitious and systematic contributions published on this topic to date and, for this reason, deserves the attention of anyone working or training in this field.
What Problems Are the Authors Attempting to Address?
Joseph et al. (2026) begin with a reality familiar to all mental health professionals: the discrepancy between the growing demand for psychological support and the limited availability of services. The World Economic Forum estimates that approximately 85% of individuals experiencing mental health difficulties receive no treatment at all, with disparities reaching up to fortyfold differences between high-income and low-income countries (Fowler & Lester, 2024).
Within this context, as the authors themselves observe, SST has already emerged as a scalable and efficient response. Many clients spontaneously seek this format, and research demonstrates that even a single encounter may produce significant and lasting change (Talmon, 1990; Cannistrà & Piccirilli, 2021; Schleider et al., 2025). The question the authors raise is therefore the following:
“Can AI further expand this reach without compromising the quality and ethical integrity of the intervention?”
What Do We Already Know About AI in Mental Health?
Before presenting their framework, the authors conducted a synthesis of sixteen studies published between 2021 and 2025 on the use of AI in mental health care. The resulting picture is rich, though not without contradictions.
What Works
Applications of AI in mental health generally fall into two broad categories: on the one hand, chatbots designed directly for users (such as Woebot, Wysa, Tessa, and Vivibot); on the other, systems designed to support professionals by analyzing clinical data, providing real-time insights, or automating administrative tasks. Literature reviews report significant reductions in symptoms of depression, anxiety, and stress, as well as improvements in constructs such as self-compassion and resilience (Casu et al., 2024; Mansoor et al., 2025; Thakkar et al., 2024).
One of the most frequently cited advantages is unlimited accessibility. Chatbots are available 24/7, do not require appointments, reduce geographical barriers, and provide an environment perceived as less judgmental, thereby facilitating disclosure among adolescents and individuals concerned about stigma (Le Glaz et al., 2021).
What Still Does Not Work
Current AI models continue to present risks related to errors and algorithmic bias, particularly when training data fail to represent culturally and linguistically diverse populations. The lack of transparency in AI decision-making processes also makes it difficult for clinicians to assess their reliability. Furthermore, important ethical concerns remain regarding data privacy, informed consent, accountability in cases of error, and the risk of social isolation among individuals who may develop dependency on relationships with chatbots (Koutsouleris et al., 2022; Algumaei et al., 2025).
One of the most debated limitations concerns empathy. As the authors observe, mental health professionals generally believe that AI will never be able to replicate human empathic understanding, which remains the core foundation of effective psychotherapy (Boucher et al., 2021). It is precisely this awareness that guides the proposal of Joseph et al. (2026), whose aim is to position AI as a support to human practitioners rather than as their replacement.
Why Is SST the Ideal Context for Integration with AI?
The authors dedicate an entire section to explaining why SST is, among therapeutic approaches, particularly well suited for integration with AI. Their reasons are both theoretical and pragmatic.
From a theoretical perspective, the focused, goal-oriented, and time-limited structure of SST naturally aligns with the characteristics of AI-based interventions, which are similarly brief, task-specific, and solution-oriented.
As Eliot argued in a 2024 analysis published by Forbes, SST and AI appear, in some ways, to be “designed” to converge. Research by Joseph and Rajan (2024) on the bibliometric growth of SST over the last decade further supports this interpretation, highlighting the increasing demand for psychological support and the need for scalable tools capable of meeting it.
From a pragmatic perspective, SST has already been successfully implemented in low-resource and high-demand settings, including disaster response, humanitarian contexts, migrant support services, and school environments (Cannistrà & Piccirilli, 2021). These are precisely the contexts in which AI may play a crucial role in the training of non-specialized practitioners.
Finally, there is already preliminary evidence supporting the feasibility of such integration. Vowels et al. (2024) demonstrated that a Large Language Model (LLM) such as ChatGPT-4 is capable of conducting realistic single-session interventions addressing relational difficulties, with users reporting positive evaluations regarding usefulness and perceived empathy. Athanassopoulos (2026) further showed how Natural Language Processing (NLP) may identify specific therapeutic markers and provide scalable feedback to counselling trainees.
Conclusion
The picture emerging from the review conducted by Joseph et al. (2026) is that of a rapidly evolving field that nevertheless remains marked by complexity and contradiction. These are tools that appear effective on certain measurable outcomes, yet continue to struggle with deeper questions concerning empathy, responsibility, and the quality of the therapeutic relationship. It is precisely from these reflections that the authors developed their proposal.
In the second article, we will examine how Joseph et al. (2026) translate this awareness into a concrete framework composed of three distinct modules in which AI supports SST in training, supervision, and even during the session itself, in real time.
Angelica Giannetti
Psychologist, Psychotherapist
Team Member, Italian Center for Single Session Therapy
Bibliography
Algumaei, A., Yaacob, N. M., Doheir, M., Al-Andoli, M. N., & Algumaie, M. (2025). Symmetric therapeutic frameworks and ethical dimensions in AI-based mental health chatbots (2020–2025): a systematic review of design patterns, cultural balance, and structural symmetry. Symmetry, 17(7), 1082.
Athanassopoulos, L. (2026). Can AI train the next generation of counsellors? Ethical challenges and opportunities. Journal of Psychology and AI, 2(1), 2606673.
Boucher, E. M., Harake, N. R., Ward, H. E., Stoeckl, S. E., Vargas, J., Minkel, J. & Zilca, R. (2021). Artificially intelligent chatbots in digital mental health interventions: a review. Expert review of medical devices, 18(sup1), 37-49.
Cannistrà, F. (2024). How to maximize a single session? From a reflection on theoretical pluralism to the definition of five goals. Journal of Systemic Therapies, 43(1), 105–120.
Cannistrà, F., & Piccirilli, F. (2021). Single-session therapy: Principles and practice (1a ed.). Giunti Psychometrics.
Casu, M., Triscari, S., Battiato, S., Guarnera, L., & Caponnetto, P. (2024). AI chatbots for mental health: A scoping review of effectiveness, feasibility, and applications. Applied Sciences, 14(13), 5889.
Eliot, L. B. (2024). Generative AI and the great promise of single-session therapy for mental health. Forbes, 3 maggio 2024.
Fowler, H., & Lester, J. (2024, November 1). How AI could expand and improve access to mental health treatment. World Economic Forum.
Joseph, J., Pai, M., Asai, K., & Turek, M. (2026). AI-Human synergy in single-session therapy: Augmenting capacity for scalable mental health support. Journal of Psychology and AI, 2(1), 2636264.
Joseph, J., & Rajan, S. K. (2024). Evolution of single-session therapy: A bibliometric analysis. American Journal of Psychotherapy, 77(2), 71–78.
Koutsouleris, N. et al. (2022). From promise to practice: Towards the realisation of AI-informed mental health care. Lancet Digital Health, 4(11), e829–e840.
Mansoor, M., Hamide, A., & Tran, T. (2025). Conversational AI in pediatric mental health: A narrative review. Children, 12(3), 359.
Schleider, J. L. et al. (2025). Single-session interventions for mental health problems and service engagement: Umbrella review. Annual Review of Clinical Psychology, 21(1), 279–303.
Talmon, M. (1990). Single-session therapy: Maximizing the effect of the first (and often only) therapeutic encounter. Jossey-Bass.
Thakkar, A., Gupta, A., & De Sousa, A. (2024). Artificial intelligence in positive mental health: A narrative review. Frontiers in Digital Health, 6, 1280235.
Vowels, L. M. et al. (2024). AI in relationship counselling: Evaluating ChatGPT’s therapeutic capabilities. Computers in Human Behavior Artificial Humans, 2(2), 100078.
WHO (2021). Comprehensive mental health action plan 2013–2030. World Health Organization.