Responsible AI Integration in a Rapidly Changing Landscape

Guest Blog Post by Kelly O’Horo, LPC and Patrick O’Horo, LAC

Kelly O’Horo

Patrick O’Horo

Artificial Intelligence (AI) is becoming an everyday presence, and mental health care is no exception. As EMDR therapists who have presented this topic at national conferences, we have seen both the great promise and the potential pitfalls of AI in our field. We believe AI can streamline aspects of therapy (like notetaking or psychoeducation), but it can never replace the human connection and clinical wisdom that effective EMDR therapy requires. In this post, we attempt to answer key questions about integrating AI into EMDR and mental health practice – sharing what we have learned from using AI in our own trauma therapy work and highlighting what current research says.  

Our background and interest in AI and EMDR therapy

We come to this topic from combined backgrounds in technology and trauma therapy. Patrick’s career has been in software and AI for nearly 20 years at Microsoft, and now includes being a licensed therapist, and Kelly is an EMDR therapist and EMDRIA Approved Trainer™ who leads a trauma-focused group practice. Together, we saw an opportunity to bridge human expertise with technology and even co-developed an electronic health record system with built-in AI tools. Our interest in AI grew out of practical needs: we were overwhelmed by paperwork and admin tasks and knew AI could help with administrative tasks, data reviews and analysis, and documentation demands.   

We also quickly recognized AI’s limits and risks. These tools can misunderstand context, make mistakes, or raise privacy concerns. We remain determined to use AI responsibly. The bottom line is that technology should support our work, not supplant human relationships and judgment at its core.  

Can AI replace human EMDR therapists?  

In a word, no. While AI keeps improving, it cannot replicate a human EMDR therapist. The essence of EMDR is the attuned therapeutic relationship, and no algorithm has real empathy or intuition. We often remind people that AI might know the definition of “sad,” but “there actually is no compassion or empathy; it’s just programming.” A machine cannot truly sit with someone’s pain, nor can it tailor its approach on the fly like a human therapist can.  

That said, AI can serve as a powerful assistant. It is exceptionally good at digesting large amounts of information and finding patterns. For example, AI might scan many pages of progress notes (considering any HIPAA and confidentiality requirements) and highlight recurring themes or subtle indicators that a busy therapist could miss. In these narrow, data-driven tasks, AI’s speed and consistency can augment a clinician’s work.  

However, any benefits come with serious limits. AI sometimes makes mistakes or even fabricates information (“hallucinates”) when it tries to predict answers. It has no genuine understanding of context, no sense of ethics, and no ability to truly connect. AI will not automatically know if its advice is dangerous or inappropriate. It can’t push back when a client is in denial or deeply hurting. That is why a trained clinician must always stay in control. We double-check and edit all AI-generated content, and we never let the machine make decisions for us. In our view, “AI should help you do your job, not do it for you.” Human therapists – with their capacity for empathy, creativity, and situational judgment – are irreplaceable in EMDR Therapy.  

The role of chatbots in therapy

AI-driven chatbots and virtual “therapy” apps are now widely available, and some people turn to them for quick support. While these bots can offer a friendly ear and basic guidance at any hour, a chatbot is not a human therapist, and chatting with an AI chatbot is not the same as real therapy.  

The core limitation is that chatbots lack true understanding and responsibility. They might sound caring as they’re designed to use sympathetic language, but they simply generate responses based on patterns, not genuine empathy. As we warn colleagues, a bot may “appear to think and be creative… They’re not. They’re just following instructions.” A chatbot may or may not recognize an emergency or provide the kind of accountability a human can. An APA health advisory reports that millions of people have tried using AI chatbots to discuss serious issues like suicidal thoughts, which is alarming because no bot can conduct a real risk assessment or intervene in a crisis (APA, 2025).  

 Chatbots might work as adjuncts to therapy, perhaps guided self-help exercises, psychoeducation, or to send appointment reminders, but they must be used with caution. If a client uses a mental health bot, a human clinician should be in the loop to monitor and interpret what is happening. Ultimately, chatbots can supplement certain aspects of mental health care, but they cannot replace the human connection, intuition, and ethical accountability of a live therapist.  

How EMDR therapists are using AI (and best practices)

EMDR therapists are finding that AI can help with many time-consuming tasks. Popular uses include documentation (AI transcription and drafting of session notes), content creation (writing psychoeducational handouts or generating website images), and other routine chores. For instance, a well-tuned AI can produce a solid first draft of a progress note within seconds, but it still falls to the therapist to review and refine that note with their own clinical impressions.  

When integrating AI, some best practices are essential: 

  • Protect confidentiality: never enter private client information into a public AI service, which might store or learn from that data. Use only HIPAA-compliant or enterprise solutions that keep data secure (for example, we use a vetted, closed-platform HIPPA compliant AI instead of open consumer chatbots).  
  • Be transparent. If you use AI for your notes inform clients how you’re using AI and obtain their consent in plain language before doing so.  
  • Another rule is to keep a human in the loop. Treat AI like a junior assistant, it can handle paperwork and suggest ideas, but you the human remain the decision-maker.  
  • Always double-check AI outputs for accuracy and appropriateness and add the nuance that only human professionals can provide.  

Used wisely, AI can lighten your load and even improve quality. In our practice, using an AI note-taker has made documentation more efficient and let us be more present with clients – in Kelly’s words, “I don’t have to worry about jotting it all down… it’s got it, and it’s got it better than I would have had it. My notes have never been better.” But we must remain vigilant: the therapist is ultimately responsible for the work, and no AI should ever override your professional judgment.  

When clients bring AI into sessions

Many clients are experimenting with AI for mental health, asking chatbots for advice or using self-help apps, so it’s important to talk about these experiences in therapy. A major concern is accuracy: generative AIs often sound confident but can give misinformation or overly simplistic answers. We encourage clients to take AI advice with a healthy dose of skepticism and to bring up anything AI tells them in session so we can discuss and fact-check it together. Together, we can examine whether the bot’s suggestions are truly helpful and evidence-based, or if they might be misleading.  

We also help clients avoid over-reliance on AI. If someone becomes attached to an always-agreeable AI “friend,” we explore what need that is fulfilling and how it differs from human support. We point out that a bot will never challenge distorted thoughts or set boundaries – it exists to satisfy users and keep them engaged. As we observed in our own practice, an AI response has “the desire… to be your friend and to make you happy,” whereas a real therapist sometimes must tell you things you don’t want to hear (but need to). By highlighting these differences, we help clients recognize that while it’s fine to use AI as a supplemental tool, it can’t replace the honest feedback and personal connection of human therapy.

Impact of AI on attachment and trust

EMDR therapy depends on a safe, attuned relationship, so the rise of AI “companions” raises important questions. Some clients have formed strong emotional bonds with AI chatbots, even saying the bot understands them better than people do. While it is positive that they feel heard, the quality of that connection is fundamentally different. An AI has no real emotions or needs, it is a one-way projection. Such a relationship can feel safer (the bot never gets upset or leaves), but it might also reinforce the avoidance of real human connection. We handle this topic with compassion, exploring what the AI companionship represents for the client and gently reminding them that no matter how friendly a bot is, it cannot offer the true empathy and responsiveness that a human can. In fact, research on EMDR therapy and attachment finds that a strong, positive therapeutic bond is central to effective trauma healing, a level of attunement that technology cannot provide (Leeds, 2023).  

Inside the therapy room, we also consider how our own use of AI tools might affect client trust. For example, if we use an app to help transcribe sessions or write notes, a client may worry about a “third presence” listening in. To address this, we clearly explain what the AI is doing and emphasize that it’s only a helper in the background, as well as the HIPAA compliant privacy process that the therapist is using. The client is conversing with us, not the device, and the device has no independent mind or judgment. We reassure clients by saying “you are the human in the room.” We have full control over the process, and their words remain private. Of course, if anyone feels uncomfortable, we don’t use the tool. Nothing should ever come between the client and the therapist’s ability to truly hear and understand each other.

Multicultural considerations

Cultural competence is as important with AI as in any other aspect of therapy. AI systems are not culturally neutral; they learn from data that may underrepresent certain groups or carry biases. An AI model “doesn’t know what it’s like to be African American… or what it’s like to be adopted.” Lacking lived experience, AI may inadvertently reinforce stereotypes or overlook cultural nuances.  

What’s striking is that even outside psychotherapy, clinicians describe the same “human” work that is deeply shaped by values, identity, and context: a JMIR qualitative study of UK general practitioners’ views on AI, participants emphasized that care often requires “listening to ideas, concerns and expectations and negotiating a shared plan,” and that “different patients have different views,” making patient management an ongoing process of negotiation rather than a one-size-fits-all decision tree. They also noted that technology is likely to struggle with individualized, nonverbal cues and the shifting way humans describe symptoms over time, and raised concerns about whether new AI-driven care models will be acceptable to patients (Blease et al., 2019).  

In therapy, those are exactly the places where culture and identity live: in meaning-making, nonverbal communication, trust, and what “help” is allowed to look like. That’s why therapists must use their own cultural awareness to filter and correct AI outputs. If an AI-generated suggestion doesn’t fit a client’s cultural context, language, lived experience, or values, we modify or discard it. We also advocate for the development of bias-aware AI and the importance of pertinent ethical considerations around bias and fairness, transparency and accountability, and the need for ongoing evaluation and safeguards. 

By staying attuned to each client’s culture and identity, we can ensure that AI serves everyone respectfully, rather than exacerbating disparities.  

Recent applicable research on AI and mental health

Source (Year)   Focus   Key Takeaway  
Doocnaik (2026)   AI in mental health care (scoping review)   AI could expand access and aid diagnosis/treatment, but risks (like over-reliance) need more attention – benefits must be weighed against potential harms.  
Haber et al. (2024)   Generative AI as “artificial third” in therapy   AI may act as a helpful third presence in therapy but also brings new ethical and relational challenges. It should enhance – not replace – the human therapist in the therapeutic relationship.  
Held et al. (2025)   AI tools in PTSD treatment (review)   Emerging tools (e.g. VR and chatbots) show promise to support PTSD treatment (exposure, skills practice), but therapist oversight is essential due to privacy, safety, and ethical issues in trauma care.  
Perlis (2026)  Commentary on AI in mental health (JAMA)   AI can improve access, efficiency, and personalized treatment, but if misused it might reduce human contact or erode clinician skills. A balance of innovation with strong ethics, transparency, and clinician training is urged.  
Saeidnia et al. (2024)   Ethical frameworks for mental health AI   International guidelines emphasize privacy, transparency, fairness, and accountability in mental health AI. Reducing bias and keeping human oversight in any AI-assisted clinical decision-making are top priorities.  
Driessen et al. (2024)   EMDR in medical settings (systematic review)   EMDR has shown efficacy in various health conditions beyond PTSD, underscoring the importance of the therapist’s flexibility and case-by-case judgment – qualities of human care that a one-size-fits-all AI cannot easily emulate.  
Leeds (2023)   Positive affect & attachment in EMDR   Highlights the central role of interpersonal positive affect and attunement in successful EMDR therapy. Technology should never disrupt the therapeutic alliance, which is vital for trauma processing and client stability.  

Free resources for learning about AI and ethics

There are many places to learn more, but some starting ideas would be:

Final thoughts

We approach AI with cautious optimism. In our own practice, we’ve seen how it can reduce paperwork and even provide new clinical insights, but we remain aware of its dangers if used carelessly. Our guiding principle, as we often say, is that “technology must support, not disrupt, the counselor–client relationship.” We should use AI to enhance our work (not to do the actual counseling), and be ready to set it aside whenever it doesn’t serve a client’s needs. No matter how advanced AI becomes, it will never replace the healing power of human empathy, wisdom, and the therapeutic bond. By keeping that truth at the center of our practice, we can safely harness AI’s benefits while preserving what matters most in EMDR therapy: the human connection.  

 

Kelly O’Horo, LPC, leads Infinite Learning as an EMDRIA Approved Consultant™ and Approved Trainer™ supporting clinicians worldwide in practical, ethical application. She co-founded Therapist Genie, an AI integrated EHR built to streamline documentation while protecting clinical judgment. Kelly founded Infinite Healing & Wellness and hosts Adaptable Behavior Explained, sharing mental health and growth tools.

Patrick O’Horo, LAC, is a Principal Technologist at Microsoft with 19 years advising Fortune 500 leaders on generative AI, Microsoft 365 Copilot, and modern work adoption. He delivers executive briefings and workshops for executives and teams, translating complex AI into practical, human centered decisions. EMDR therapist; co-creator of Therapist Genie.

 

References 

Abrams, Z. (2025, March 12). Using generic AI chatbots for mental health support: A dangerous trend. American Psychological Association Services. https://www.apaservices.org/practice/business/technology/artificial-intelligence-chatbots-therapists

AI Mental Health Collective. (n.d.). Resources. Retrieved March 23, 2026, from https://www.aimentalhealthcollective.com/resources

American Psychological Association. (2025). AI practice management: Practical guidance for psychologists. https://www.apa.org/pubs/reports/practitioner/2025/ai-practice-management

American Psychological Association. (2025). APA health advisory on the use of generative AI chatbots and wellness applications for mental health. https://www.apa.org/topics/artificial-intelligence-machine-learning/health-advisory-chatbots-wellness-apps

Blease, C., Kaptchuk, T. J., Bernstein, M. H., Mandl, K. D., Halamka, J. D., & DesRoches, C. M. (2019). Artificial intelligence and the future of primary care: Exploratory qualitative study of UK general practitioners’ views. Journal of Medical Internet Research, 21(3):e12802.  https://doi.org/10.2196/12802  

Docnaik, A. (2026). Artificial Intelligence and the Potential Transformation of Mental Health – Real-Time AI Scoping Review. Psychiatry AI. https://psychiatryai.com/artificial-intelligence-and-the-potential-transformation-of-mental-health/

Driessen, E., Morsink, S., Busschbach, J. J. V., Hoogendijk, W. J. G., & Kranenburg, L. W. (2024). EMDR treatment in the medical settings: A systematic review. European Journal of Psychotraumatology, 15(1), 2341577. https://doi.org/10.1080/20008066.2024.2341577

Haber, Y., Levkovich, I., Hadar-Shoval, D., & Elyoseph, Z. (2024). The artificial third: A broad view of the effects of introducing generative artificial intelligence on psychotherapy. JMIR Mental Health, 11, e54781. https://doi.org/10.2196/54781

Held, P., Statde, E. C., Dondanville, K.,  &Wiltsey Stirman, S. (2025). Generative artificial intelligence in posttraumatic stress disorder treatment: Exploring five different use cases. Journal of Traumatic Stress, 38(5), 813-820.  https://doi.org/10.1002/jts.23188

Kantayya, S. (Director). (2020). Coded bias [Film]. 7th Empire Media. https://www.codedbias.com/

Leeds, A. M. (2023). Foundations of the positive affect tolerance protocol: The central role of interpersonal positive affect in attachment and self-regulation. Journal of EMDR Practice and Research, 17(3), 139–158. https://doi.org/10.1891/emdr-2023-0006

Perlis, R. H. (2026). Artificial intelligence and the potential transformation of mental health. JAMA Psychiatry, 83(4), 409–413. https://doi.org/10.1001/jamapsychiatry.2025.4116 

Saeidnia, H. R., Hashemi Fotami, S. G., Lund, B., & Ghiasi, N. (2024). Ethical considerations in artificial intelligence interventions for mental health and well-being: Ensuring responsible implementation and impact. Social Sciences, 13(7), 381. https://doi.org/10.3390/socsci13070381 

 

Back to Focal Point Blog Homepage

 

Additional Resources

If you are a therapist interested in the EMDR training:

If you are EMDR trained:

If you are an EMDRIA™ Member:

 

Date
April 24, 2026

Contributor(s)
Kelly O'Horo, Patrick O'Horo

Practice & Methods
Virtual Reality/Technology, Your EMDR Practice

Audience
EMDR Therapists, General/Public

Language
English

Content Type
Blog/Blog Post

Access Type
Open Access

Related Articles