The American Psychological Association has released AI Essentials for Practitioners, five short video tutorials that show clinicians practical, low-risk ways to use free AI tools in everyday practice. One of the five is led by Sentio co-founder and Chief Academic Officer Alexandre (Alex) Vaz, PhD: “How to verify AI outputs: A 3-step accuracy check for clinicians.”
The other four tutorials, presented by Drs. David Cooper, Amber Childs, Charmain Jackman and Kay Nikiforova, cover writing better prompts, creating psychoeducational handouts, writing a professional bio and using AI for literature reviews. APA offers the series as a first step for clinicians who are curious about AI, with ethics and clinical judgment kept in view throughout.
Check, Verify, Revise
AI tools can produce answers that sound authoritative and are still wrong. Dr. Vaz demonstrates this with a realistic case. A clinician asks ChatGPT what the research says about narrative exposure therapy for complex PTSD in war veterans, planning to use the answer on her practice website. The reply names a specific study in a specific journal, reports that more than 80 percent of veterans reached full remission within 20 weeks, and states that the World Health Organization has adopted the protocol as a first-line treatment worldwide. When she consults her textbooks, colleagues and official guidelines, she finds that the study was fabricated and the journal does not exist.
The tutorial walks through three steps for catching errors like these before they reach clients or a website:
- Check the output. Look for vague statements, sweeping conclusions, claims with little nuance or no mention of limitations, and anything that sounds authoritative but cites no sources.
- Verify what matters. Test the key claims against authoritative sources, such as official practice guidelines, and against your own professional knowledge. Asking the AI for the sources behind its claims can expose problems too. In the example, it could not find any.
- Revise before using. Remove or soften claims that could not be confirmed, keep what was verified, and rewrite the text with your own clinical judgment and in your own voice.
The revised website text in the example makes a more modest claim: research suggests that narrative exposure therapy can reduce PTSD symptoms, although outcomes vary. Dr. Vaz notes that AI gets a great deal right, and that every output still needs all three steps before it is used.
Watch the tutorial on the APA website
Part of Sentio’s Work on Safe and Ethical AI
Dr. Vaz opens the tutorial by noting that Sentio was among the first programs to build AI training for its students, faculty and supervisors. That work includes the Sentio AI research program, which studies how people use large language models for mental health support and how safely those systems respond; a free AI course for mental health professionals; clinical supervisor training that covers the safe and ethical use of AI; and A Clinician’s Guide to AI-Supported Therapy Homework, the new APA book by Dr. Vaz and Tony Rousmaniere. The Sentio MFT program brings the same approach to the training of future therapists, guided by the Sentio Statement on AI.
The tutorial closes by pointing viewers to the rest of the AI Essentials series and to APA’s ethical guidance for AI in professional practice.
About this video
- Title
- How to verify AI outputs: A 3-step accuracy check for clinicians
- Presenter
- Alexandre (Alex) Vaz, PhD, Sentio co-founder and Chief Academic Officer
- Series
- AI Essentials for Practitioners, five short tutorials
- Publisher
- American Psychological Association
- Released
- November 2026
- Length
- About 7 minutes
- Watch
- AI Essentials for Practitioners on the APA website
More from Sentio on AI and mental health: the AI research program, the free AI course for therapists, the Sentio Statement on AI and the Sentio MFT program.
