What helps and what hinders the adoption of AI in psychotherapy? Focus groups with patients and therapists

Sat, 07 Mar 2026 00:00:00 +0000

AI technologies could reduce therapists’ workload and improve mental healthcare – yet their adoption in psychotherapy remains challenging. In this study led by Julia Cecil (with me as senior author), we asked those who would actually work with these tools: What would make patients and therapists adopt AI in psychotherapy, and what would hold them back?

What we did

We conducted six online focus groups with a total of 32 participants in Germany – 19 therapists (mostly in training) and 13 patients in psychotherapy for depression or anxiety. The discussion guide was based on the NASSS framework, and we analyzed the data with a combined deductive and inductive thematic analysis.

What we found

Across the seven NASSS domains, 36 categories emerged: 16 facilitators (e.g., useful technology elements, customization to users’ needs, cost coverage), 11 barriers (e.g., lack of human contact, resource constraints, AI dependency), and 9 factors that could work either way depending on the context (e.g., the therapeutic approach, institutional differences).
Nearly half of all statements concerned the technology itself and the people who would use it.
Patients and therapists raised largely the same topics. Participants saw AI as more suitable for anxiety and depression than for conditions such as psychosis, and as more compatible with structured approaches like CBT than with psychodynamic therapy.

Categories regarding AI adoption in psychotherapy, clustered by NASSS domain (blue = facilitator, red = barrier, yellow = mixed factor; numbers = number of codes). Figure 2 from Cecil et al. (2026), npj Mental Health Research, licensed under CC BY 4.0.

What this means

Adopting AI in psychotherapy is complex, and the same factor can help or hinder depending on the setting. Developers should involve patients and therapists early and address barriers – above all, preserving the human relationship at the core of therapy – from the start of development.

Read the paper

Cecil, J., Schaffernak, I., Evangelou, D., Lermer, E., Gaube, S., & Kleine, A.-K. (2026). Navigating the complexity of AI adoption in psychotherapy by identifying key facilitators and barriers. npj Mental Health Research, 5, 17. https://doi.org/10.1038/s44184-026-00199-1

Why AI decision support in ophthalmology is approved but rarely used: An interview study

Wed, 22 Oct 2025 00:00:00 +0000

AI could ease the growing workload in ophthalmology, and several AI-enabled clinical decision support systems (AI-CDSS) have already received regulatory approval. Yet they are rarely used in practice. In this preregistered study led by Insa Schaffernak, we asked what it takes for ophthalmology professionals to adopt these tools and to keep using them.

What we did

We conducted semi-structured interviews with 22 ophthalmology professionals from Germany, Austria, and Switzerland – residents, attending physicians, heads of department, assistants, optometrists, and study nurses – with very different levels of experience with AI. We analyzed the interviews with a qualitative content analysis guided by the NASSS framework (Nonadoption, Abandonment, Scale-up, Spread, and Sustainability).

What we found

Most participants were open to AI, especially as a complement to their own expertise – as a second opinion for complex cases, or as support for easy but repetitive assessments where fatigue can lead to oversights.
Whether AI is adopted and adds value depends on 29 sociotechnical factors across all seven NASSS domains, from usability and validated performance to workflows, costs, regulation, and patients’ perceptions.
Adoption is not a one-time decision: clinicians re-evaluate AI-CDSS repeatedly, weighing expected or experienced value against costs – both before adopting a tool and afterwards, when deciding whether to continue or abandon it.
Many challenges are shared with traditional health technologies. What sets AI apart is how users psychologically appraise it.

Sociotechnical factors identified in the interviews and their role in adoption, continued use, or abandonment. Figure 1 from Schaffernak et al. (2025), BMC Health Services Research, licensed under CC BY 4.0.

What this means

Regulatory approval is not enough. Developers and implementers need to address users’ needs across the whole sociotechnical system – and support clinicians in evaluating the added value of AI over time. The study also shows that the NASSS framework is a useful lens for studying AI implementation.

Read the paper

Schaffernak, I., Cecil, J., Kleine, A.-K., & Lermer, E. (2025). Sociotechnical influences on the adoption and use of AI-enabled clinical decision support systems in ophthalmology: A theory-based interview study. BMC Health Services Research, 25, 1398. https://doi.org/10.1186/s12913-025-13620-w

qualitative research on Anne-Kathrin Kleine

What helps and what hinders the adoption of AI in psychotherapy? Focus groups with patients and therapists

What we did

What we found

What this means

Read the paper

Why AI decision support in ophthalmology is approved but rarely used: An interview study

What we did

What we found

What this means

Read the paper