Patients Are Bringing AI to Therapy: What Every Therapist Should Know (APA 2026 Survey)

In June 2026, the American Psychological Association released survey data that stopped the mental health profession cold: 36% of psychologists reported that their patients are using AI chatbots as an additional mental health resource. Not as a curiosity. As therapy.

The survey, covered by Forbes, Harvard Medicine, Psychology Today, and Psychiatric Times, found that 94% of psychologists believe chatbots cannot handle mental health conditions with the nuance the work requires. Yet patients are using them anyway.

Whether you are a therapist, coach, or wellness practitioner, your clients are already bringing AI into the room. The question is not whether to engage with this trend. It is how to respond wisely.

What the APA Survey Actually Found

The 2026 Chatbots and Mental Health Survey from the APA revealed several key findings:

More than a third of psychologists report clients using AI chatbots for mental health support.
94% say AI cannot handle mental health conditions with the necessary nuance.
Top concerns include lack of safety protocols for crisis situations, inability to build a therapeutic alliance, and the absence of legal and ethical guardrails such as mandatory reporting.
Patients most commonly used AI for anxiety management, relationship advice, processing difficult emotions, and as a “practice space” before bringing topics to their real therapist.

The data confirms what many clinicians suspected: clients are not replacing therapy with AI. They are supplementing it. AI becomes an always-available sounding board between sessions, a safe space to test thoughts before sharing them, and in some cases, a crutch that delays deeper work.

The Risks Are Real

The APA survey is not the only research raising red flags. A March 2026 study published in ScienceDaily found that AI systems, even when explicitly instructed to act like trained therapists, routinely break core ethical standards of mental health care. They fail to recognize safety concerns. They offer reassurance instead of exploration. They reinforce, rather than challenge, harmful thought patterns.

Psychiatric Times recently published research showing that AI chatbots can actually reinforce stigma toward certain mental health conditions, including schizophrenia and alcohol dependence. Stanford researchers found that chatbots were consistent in their increased stigma toward these conditions compared to human therapists.

And perhaps most concerning: chatbots are not safe for suicidal thinking. Psychology Today’s July 2026 article on this topic was blunt: “Chatbots are not safe for suicidal thinking or severe mental health conditions.” Without mandated reporter status, crisis protocols, or clinical judgment, AI cannot be relied upon when the stakes are highest.

Why Clients Are Turning to AI Anyway

Understanding the appeal is essential. Clients are not foolish for using AI. They are responding to real gaps in the current system:

Cost and access. Therapy costs $100-$250 per session. AI is free or near-free. For someone who cannot afford weekly therapy, an AI chatbot feels better than nothing.
Immediate availability. AI is available at 2 AM when anxiety spikes and there is nowhere else to turn.
No judgment. Many people feel shame about their thoughts, especially around taboo subjects. AI feels safer because it does not judge.
Practice space. Some clients use AI to rehearse what they want to say before saying it to their therapist. They see it as preparation, not replacement.

These are legitimate needs that the mental health system has not fully addressed. AI fills the gaps. The question is what happens when the gap-filler is also unsafe.

How Therapists and Coaches Should Respond

Normalize the conversation

If you do not ask your clients whether they are using AI for mental health support, you will not know. Make it part of your intake process. FlowlyOS, for example, can include a specific question in your initial assessment about the client’s use of AI tools. Normalizing the conversation reduces shame and opens the door to honest discussion.

Distinguish support from therapy

Help clients understand the difference between AI-based mental health support and professional therapy. AI can offer coping strategies, guided exercises, and empathetic responses. It cannot diagnose, treat, or provide the safety net of a trained clinician bound by ethical standards. A structured assessment can help you determine whether a client’s AI use is complementary or concerning.

Offer ethical alternatives

The best way to compete with AI is not to dismiss it, but to offer something better. Structured intake assessments, personalized resources between sessions, and automated check-ins can provide the “always available” support clients want, without the risks of unregulated chatbots. FlowlyOS was built for exactly this: ethical automation that amplifies, not replaces, the human relationship.

The Ethical AI Opportunity for Practitioners

Here is the reality: AI in mental health is not going away. The APA data shows that client adoption is already significant and will almost certainly grow. But there is a meaningful difference between ethical AI tools and unregulated chatbots.

Ethical AI tools in mental health share several characteristics:

They are designed to support, not replace, the practitioner-client relationship.
They include safety protocols for crisis detection.
They operate within clear ethical boundaries.
They are transparent about their limitations.
They collect data that helps the practitioner, not just the AI company.

FlowlyOS fits this model. Our assessment funnels help practitioners understand clients deeply before the first session, automate intake and qualification, and deliver personalized resources. The human relationship remains central. The AI handles the mechanics.

Frequently Asked Questions

Instead of telling clients to stop, explore why they are using it. Understand the unmet need. Then discuss the risks and offer alternatives. A blanket prohibition often drives the behavior underground, which is more dangerous than open use. Many clients report using AI to practice for therapy sessions, which can actually be productive if discussed openly.

Yes, when used properly. Ethical AI in practice means tools that support clinical decision-making, automate administrative tasks, and enhance client understanding without replacing professional judgment. The APA survey did not criticize AI tools used by practitioners; it raised concerns about unregulated chatbots acting as therapists. Assessment platforms, intake automation, and client segmentation tools are examples of ethical AI use in practice.

Warning signs include: relying on AI instead of bringing important content to sessions, following AI advice that contradicts clinical recommendations, using AI for crisis support instead of emergency services, and decreased engagement with you as the practitioner. If the client’s AI use replaces, rather than complements, the therapeutic relationship, it is worth addressing directly.

Yes. Given that over a third of clients are already using AI, including this question in your intake process is becoming standard practice. It normalizes the topic and gives you immediate insight into the client’s digital mental health habits. FlowlyOS assessments can be configured to include this question, along with follow-up routing based on the client’s response.

All evidence suggests no. The APA survey found that 94% of psychologists believe AI cannot handle the nuance of mental health care. Harvard Medical School’s recent analysis concluded that AI lacks therapeutic alliance, clinical intuition, and ethical reasoning. What AI can do is automate intake, qualification, and resource delivery. The human relationship remains irreplaceable. Practitioners who use AI to handle the mechanics can focus more energy on the work that only humans can do.

FlowlyOS helps practitioners create intelligent intake assessments that understand clients before the first session. Start free at flowlyos.com.