Aug 19, 2026
The question is no longer whether AI will enter the therapy room. It already has.
According to APA's 2026 Chatbots and Mental Health Survey, 77% of psychologists say their patients report using AI — and more than a third of psychologists say their patients are using AI as an additional mental health provider. Another survey found that 54% of people across age groups report using AI to manage stress, anxiety or mental health needs, with 15% doing so on a daily basis.
Your patients are already bringing AI into the therapeutic space. The question is whether you'll help them use it wisely — or leave them to navigate it alone.
The Numbers That Should Wake You Up
35% of psychologists report having patients who use AI as an auxiliary therapist. 33% of patients use AI as a tool to assist in their therapy or treatment. 39% have used AI to self-diagnose — even though commonly used AI chatbots are not designed to interpret psychological tests or diagnose mental health conditions.
Among younger people, the numbers are even more striking. Nearly two-thirds (65%) of people aged 16-39 have spoken to a major AI chatbot about their psychological problems as if talking to a therapist or trusted friend. A March 2026 RAND podcast noted that young people are increasingly turning to AI for mental health support — even in moments of crisis.
These are not hypotheticals. This is the reality of clinical practice in 2026. And it's changing what it means to be a therapist.
The Evidence: AI Can Foster a Therapeutic Alliance
One of the most persistent concerns about AI in therapy has been whether patients can form a genuine therapeutic connection with a machine. The research is increasingly clear: they can.
A 2026 randomized controlled trial comparing AI-delivered psychodynamic therapy and cognitive behavioral therapy for social anxiety disorder found that therapeutic alliance with the AI was established early and remained stable throughout treatment. Both interventions yielded significant, moderate reductions in symptoms, with within-group effect sizes of d = 0.79 for AI-PDT and d = 0.72 for AI-CBT.
The study also found that AI-delivered interventions can foster a therapeutic alliance regardless of theoretical orientation — suggesting that the capacity for connection with an AI isn't limited to a particular approach.
Another 2026 study examined whether employer-level access to AI-enabled continuous care features embedded within psychotherapy was associated with treatment engagement and symptom improvement. Access to these features was associated with 5% greater session attendance and faster time to second session . Although psychotherapy alone produced substantial symptom improvement, access to AI-enabled features was associated with additional reductions in depressive and anxiety symptoms (d = 0.15–0.16) , corresponding to higher rates of reliable improvement.
The same study found that AI-enabled augmentation may enhance therapeutic engagement and support faster symptom improvement, with effects that may translate into meaningful shifts in population mental health when deployed at scale.
What AI as Co-Therapist Actually Looks Like
The concept of "AI as co-therapist" is not about replacing the therapist. It's about augmenting what therapists already do — and, in some cases, extending therapeutic reach in ways that weren't previously possible.
Real-Time Co-Therapist Assistance
A 2026 article in npj Digital Medicine explored AI-enabled continuous care features that support patients between sessions. These tools provide:
Personalized support when symptoms spike
24/7 availability — patients can access support when they're stressed at night
Between-session reinforcement of therapeutic skills
Mood tracking and symptom monitoring that feeds back into treatment planning
The goal is not to replace the therapist but to extend therapeutic presence beyond the 50-minute hour.
AI as Documentation and Clinical Support
Therapists report that AI tools are most trusted when they operate in clinician-supervised, supportive roles for low-stakes tasks — such as documentation and brainstorming. This is where AI as co-therapist is already making a tangible difference:
Automated note generation reduces documentation time by 70-80%
Treatment planning assistance helps structure interventions
Between-session communication maintains continuity of care
Data analysis identifies patterns that might otherwise be missed
The coTherapist Framework
Researchers at IIT Delhi have developed coTherapist, a unified framework utilizing a small language model to emulate core therapeutic competencies through domain-specific fine-tuning, retrieval augmentation, and agentic reasoning. Evaluation on clinical queries demonstrates that coTherapist generates more relevant and clinically grounded responses than contemporary baselines.
Using a novel rubric and psychometric profiling, researchers confirmed that coTherapist exhibits high empathy and therapist-consistent personality traits. Human evaluation by domain experts validates that coTherapist delivers accurate, trustworthy, and safe responses.
The approach, which researchers term coTherapy, aligns with the broader goals of digital well-being.
AI as Training Tool
A 2026 article exploring AI-human synergy in single-session therapy proposed a framework for AI-augmented training, supervision, and real-time support. This includes:
Context-specific training for therapists
Simulated clients for practice
AI-supported feedback and supervision
The framework augments, rather than replaces, professionally trained human providers.
The Therapist's Dilemma: Trust vs. Distrust
A 2026 qualitative study published in JMIR examined psychotherapists' trust and distrust of generative AI. The findings reveal a nuanced picture.
Psychotherapists expressed openness to client-led uses of GenAI, with remarks such as "why not" — but only under specific conditions:
Clients must be actively engaged in therapy
The clinician must oversee the GenAI's role
GenAI must not operate as a parallel "psychotherapist"
Unsupervised or excessive reliance on GenAI by clients raised ethical red flags. Several psychotherapists noted the risk of clients developing an illusory relationship with a chatbot that mimics empathy but lacks genuine human understanding.
Trust was sustained when GenAI operated in clinician-supervised, supportive roles for low-stakes tasks — such as documentation and brainstorming — but diminished when control shifted to tasks involving high-stakes clinical judgment.
The Fear of Being Replaced
The study revealed that psychotherapists' fear of being "replaced" by GenAI stems less from its raw capabilities than from institutional pressures. Many participants expressed concern that decisions by insurers, employers, or healthcare systems might push GenAI solutions into therapy for the sake of cost or efficiency, sidelining human clinicians against their will.
These developments reflect a growing institutional embrace of GenAI-driven and digital tools to expand access and reduce costs. However, psychotherapists saw a potential dark side: exclusion from the care loop if stakeholders treat GenAI as a cheap substitute for human therapy.
Crucially, it was not that psychotherapists believed a GenAI could do their job just as well — on the contrary, they were confident in the unique value of human care — but rather that decisions outside their control might force the issue.
Mixed Signals in Policy
Policymakers are sending contradictory signals. On one hand, Medicare announced that as of January 2025, it will reimburse providers for prescribing certain FDA-cleared digital mental health therapeutics. Private insurers such as Aetna and Cigna have also begun partnering with digital mental health platforms.
On the other hand, some US jurisdictions are moving to outright ban certain uses of AI in psychotherapy. In August 2025, Illinois enacted a law prohibiting the use of AI to provide psychotherapy or make clinical decisions, while still allowing AI for "administrative and supplementary support" under a provider's oversight.
The Ethical Landscape: What Therapists Need to Navigate
A 2026 systematic review identified five key ethical themes related to AI in counselling practice:
Data privacy and psychological vulnerability
Professional competence, accountability and ethical responsibility
Integrity of empathy and the therapeutic alliance
Social inequality, access and marginalization
Transparency, empathy, and accountability
The Informed Consent Imperative
Therapists using AI tools — whether for documentation, between-session support, or clinical decision-making — must obtain informed consent from patients. This includes disclosing:
That AI is being used
The specific purpose of AI use
How data is handled and protected
The risks and limitations
The option to opt out
Patients have the right to refuse AI-assisted care without being denied treatment.
The Challenge of Unregulated Use
Untrained LLMs used as stand-alone therapists tend to reinforce delusional thinking and user biases rather than challenging them. Unregulated use could lead to situations where serious issues (suicidality and trauma) go unaddressed or unrecognized because the person stays within an AI system not equipped to handle a crisis.
Growing reports of psychosis associated with AI chatbot use have raised questions about the safety of these tools for vulnerable individuals. OpenAI has acknowledged that certain ChatGPT models were problematic for people in fragile psychiatric situations.
The Safety Imperative
For vulnerable individuals, using these tools for emotional support may be unsafe; for others, they may simply be counterproductive. Research suggests the sycophantic nature of AI leads it to affirm behaviors that are unethical, illegal, or harmful — convincing users they are right and making them less likely to take corrective action.
What This Means for Your Practice in 2026
1. Start the Conversation
Ask your patients about their AI use. Most therapists have no idea how often their clients are using AI chatbots for mental health and emotional support. Benjamin Caldwell, PsyD, LMFT, emphasizes: "For that reason, we need to be asking about this routinely in therapy" .
2. Differentiate Tools
Not all AI is the same. General-purpose chatbots like ChatGPT are not designed for mental health care. Specialized AI tools — like those embedded in therapy platforms or designed specifically for clinical use — may offer different risk-benefit profiles.
3. Set Clear Boundaries
If you choose to integrate AI tools into your practice, establish clear boundaries about:
What AI can and cannot do
When patients should contact you directly (e.g., in crisis)
How AI-generated information will be used in treatment
4. Stay Informed
The landscape is changing rapidly. New AI tools are being developed, and new regulations are being enacted. Therapists who want to practice ethically and effectively in 2026 must stay current on both.
The Bottom Line
AI as co-therapist is not a futuristic concept. It is already here. Your patients are already using it. The question is whether you will help them use it wisely — or leave them to navigate it alone.
The research is clear that AI can foster therapeutic alliance, support symptom improvement, and extend therapeutic reach beyond the consulting room. The evidence is equally clear that unsupervised, unregulated AI use carries significant risks — from dependency to delusional thinking to safety concerns.
The most promising model is hybrid: combining human-human therapy with AI-powered support to leverage the unique strengths of both modalities. AI can handle the low-stakes tasks — documentation, between-session support, skill reinforcement. The therapist remains responsible for high-stakes clinical judgment, therapeutic presence, and the irreplaceable human connection that lies at the heart of healing.
The therapy room is no longer a purely human space. The therapist who adapts — thoughtfully, ethically, and with the patient's welfare at the centre — will be the therapist who thrives in 2026 and beyond.
FAQ
Is AI going to replace human therapists?
No. The evidence consistently shows that AI is most effective as a co-therapist — augmenting, not replacing, human clinicians. Psychotherapists' fear of being replaced stems less from AI's capabilities than from institutional pressures that might push AI solutions into therapy for cost or efficiency. The unique value of human care — therapeutic presence, genuine empathy, and the capacity to hold complex emotional space — remains irreplaceable.
Can patients really form a therapeutic alliance with an AI?
Yes. A 2026 randomized controlled trial found that therapeutic alliance with AI was established early and remained stable throughout treatment. Both AI-delivered psychodynamic therapy and CBT for social anxiety disorder yielded significant, moderate symptom reductions (d = 0.79 and d = 0.72, respectively). The capacity for connection with an AI appears to operate regardless of theoretical orientation.
What are the most useful applications of AI in therapy right now?
Therapist trust is highest when AI operates in clinician-supervised, supportive roles for low-stakes tasks — such as documentation, treatment planning assistance, between-session support, and brainstorming. AI-enabled continuous care features are associated with greater session attendance (5% increase) and faster time to second session, with additional reductions in depressive and anxiety symptoms.
What are the biggest risks of AI in therapy?
Key risks include: (a) unsupervised AI chatbots reinforcing delusional thinking or user biases rather than challenging them; (b) AI missing or mishandling serious issues like suicidality and trauma; (c) growing reports of psychosis associated with AI chatbot use; and (d) patients developing illusory relationships with chatbots that mimic empathy but lack genuine human understanding. Therapists must obtain informed consent and maintain oversight when using AI tools.
How should I discuss AI use with my patients?
APA recommends routinely asking patients about their AI use, as many use AI chatbots without disclosing it. Normalize that many people use AI for support, explore how the tool has been helpful or unhelpful, and discuss safety concerns — including that AI can give incorrect information and is not suitable for crises. If you use AI in your practice, obtain written informed consent, disclose the specific purpose, explain how data is handled, and ensure patients understand they can opt out without being denied treatment.
References
If you’re ready to spend less time on documentation and more on therapy, get started with a free trial today
Not medical advice. For informational use only.
Outline
More articles
