Course curriculum

    1. Learning Orientation

    2. What exactly is an LLM doing? Understanding the Machine

    3. The Black Box

    4. Why does this matter in psychotherapy?

    5. Quiz

    1. At a Glance

    2. Setting the Scene

    3. What Does the 2026 Evidence Actually Show?

    4. The Strongest Positive Trial Signals

    5. What Do the Meta-Analyses Add?

    6. Quiz

    1. Augmentation Rather Than Replacement

    2. Safety

    3. Why 'Helpful' Is Not the Same as 'Therapeutic'

    4. User Experience, Alliance and Boundaries

    5. What Do Psychotherapists and the Public Think?

    6. Quiz

    1. Useful for certain types of supportive tasks

    2. A Clinical Reading of the Evidence

    3. Questions for Further Clinical Discussion

    4. Conclusion

    5. Quiz

    1. References

Large Language Models and Psychotherapy: What Does the 2026 Evidence Tell Us?

This reading course examines what the emerging 2026 research tells us about the use of LLMs and conversational AI in psychotherapy. The course considers what these systems can currently do, where the evidence is strongest, and where significant clinical and safety concerns remain. It provides psychotherapists with a practical framework for distinguishing between appropriate AI augmentation, functions requiring human supervision, and aspects of psychotherapy that should not yet be delegated to autonomous systems.

  • Explain the fundamental nature of large language models, including how their operation differs from human cognition and why they remain partly a “black box.”

  • Distinguish between conversational fluency and clinical competence, including the limitations of using apparent empathy, reasoning or therapeutic language as evidence of genuine understanding.

  • Interpret the emerging 2026 evidence concerning the effectiveness of LLM-based interventions, particularly in structured, CBT-like, depression- and anxiety-focused applications.

  • Identify appropriate roles for AI augmentation, including therapist support, psychoeducation, between-session assistance, training, drafting, coding and other structured tasks.

  • Recognise important clinical and ethical risks, including hallucination, over-trust, dependency, cultural limitations, privacy concerns, crisis-management failures and inadequate human oversight.

  • Apply a clinically informed framework for evaluating AI use in psychotherapy, distinguishing tasks that may be safely augmented from those requiring human judgement, supervision, accountability and therapeutic responsibility.

Presented by

CEO The Science of Psychotherapy Matthew Dahlitz

Matthew Dahlitz is both university trained and an autodidact whose knowledge spans across the arts, technology, psychology, neuroscience, emergency medicine, and business. He has studied psychology at the University of Queensland with a Master of Counselling degree specializing in neuropsychotherapy. He is the Editor-in-Chief of The Neuropsychotherapist, has taught post-graduate courses in neuropsychotherapy, and is author of the book The Psychotherapists’s Essential Guide to the Brain. Matthew has also studied at the Queensland Conservatorium of Music in both music technology and master of composition. He has scored for film and produced a number of albums of original music and played producer in a film production company. Matthew has also been an Advanced Care Paramedic for a decade, managed commercial property and property development projects in the tens of millions of dollars, founded the Age of Robots site that reports on the technology and the social and psychological impacts of this second machine age, and runs a media and publishing company with an emphasis on great story telling.

Become a Member