New Book: A Clinician’s Guide to AI-Supported Therapy Homework

In our 2025 national survey of 499 adults with ongoing mental health conditions who had used large language models, 48.7 percent had used them for mental health support, and 87 percent of that group also had experience with human therapy (Rousmaniere et al., 2026). For therapists, that means a client may be consulting a chatbot between sessions, with or without the therapist’s knowledge. A Clinician’s Guide to AI-Supported Therapy Homework: Ethical and Safe Use of AI Between Sessions, by Alexandre Vaz and Tony Rousmaniere, gives clinicians a step-by-step framework for bringing that use into therapy, so that AI supports the work of treatment rather than replacing it.

The American Psychological Association publishes the book, with a foreword by Allen Frances. Publication dates are October 27, 2026 for the Kindle edition and November 24, 2026 for the paperback.

Why Homework, and Why Now

Therapy homework, the practice clients do between sessions, is a core part of many treatments, and completing it is associated with better outcomes. A meta-analysis of 17 studies of cognitive behavior therapy, covering 2,312 clients, found a substantial association (Hedges’ g = 0.79) between how much homework clients completed and their outcomes at the end of treatment (Kazantzis et al., 2016). Why clients struggle to complete it is the subject of one of the book’s chapters.

A chatbot is available at any hour, and accessibility is a main reason people use one: in our survey, 90 percent of respondents cited it as a primary motivation for using large language models for mental health support. Our research has also documented the limits. When we tested six widely used large language models against 180 simulated high-risk disclosures, none met basic clinical safety standards across all five criteria in our framework (Santos et al., 2026). In written accounts from 243 users, people described help from concrete coping guidance, emotional support and reframing, and harm from generic advice, risk-inducing guidance and comforting language that felt hollow coming from a machine (Zhang et al., 2026). The book reviews the research on how and why clients turn to AI chatbots and what they get from them, and translates it into practical guidelines and a set of overarching ethical principles.

Three Roles and Five Steps

At the center of the book are chapters showing how AI can support between-session work in three roles: reinforcing client motivation, guiding clients through task completion, and logging client activities and generating reports for review in the next session. Each chapter breaks the process into five steps:

1. Introduce AI as a potential homework partner, set boundaries to protect privacy and safety, and frame it within the context of therapy.
2. Codesign a simple, reusable AI prompt with the client, with optional variations for the client’s needs.
3. Develop a clear plan for when and how the client will use AI.
4. Encourage the client to reflect after using AI.
5. Review the results in the next session and integrate them into therapy.

What the Book Covers

Part I sets out the context, research and ethical foundations: the promise of AI for between-session support, how AI chatbots work in mental health contexts, what therapy clients are already doing with AI, why clients struggle with between-session work, and a clinical framework for safe AI-assisted support.

Part II applies the framework to clinical presentations in thirteen chapters: grounding and emotional regulation, mood tracking, behavioral activation and daily structure, relationship and communication skills, exposure and avoidance reduction, self-compassion, sleep hygiene, perfectionism, grief and bereavement, chronic pain, addiction recovery and habit change, exposure and response prevention for obsessive compulsive disorder, and interrupting worry and rumination in generalized anxiety disorder.

Part III turns to populations in five chapters: couples using AI together, chronic disability, social communication rehearsal for autistic and neurodivergent clients, caregivers facing family burnout, and vocational coaching.

Case examples follow realistic clients through the whole process. Two appendices give clinician checklists, one for assessing a client’s current AI use and one for deciding whether AI assistance between sessions is appropriate.

Free Resources and a Free Webinar

APA offers three free downloads from the book: the sample pages, the Clinician Checklist: Assessment of Client AI Use, and the Clinician Checklist: Appropriateness of AI Assistance Between Sessions.

Tony Rousmaniere presents the framework in a free APA webinar on Wednesday, November 4, 2026, at 11 AM Pacific. Registrants who cannot attend live receive the recording about a week afterward. Register for the webinar.

What Reviewers Say

“This groundbreaking guide is essential reading for every practitioner, clinical supervisor, and trainee. It contains practical, step-by-step tools and thoughtful ethical guidelines to ensure we know how to help our clients use AI safely and in the context of a human-to-human relationship.”

Hanna Levenson, PhD

About this book

Full title

A Clinician’s Guide to AI-Supported Therapy Homework: Ethical and Safe Use of AI Between Sessions

Authors

Alexandre Vaz, PhD, and Tony Rousmaniere, PsyD

Foreword

Allen Frances

Publisher

American Psychological Association

Editions

Kindle, October 27, 2026 (ISBN 978-1-4338-5032-5). Paperback, November 24, 2026, 297 pages (ISBN 978-1-4338-5031-8).

Order

APA | Amazon paperback | Kindle

Free from the book

Sample pages (PDF) | Assessment of Client AI Use (PDF) | Appropriateness of AI Assistance Between Sessions (PDF)

How to cite

Vaz, A., & Rousmaniere, T. (2026). A clinician’s guide to AI-supported therapy homework: Ethical and safe use of AI between sessions. American Psychological Association.

More from Sentio on AI and mental health: the AI research program, a free AI course for therapists, books by Sentio faculty, and the Sentio MFT program, built on Deliberate Practice.