If you’re a therapist, you’ve probably had this moment recently: a patient mentions, almost in passing, that they talked to ChatGPT about their anxiety last week. Or you notice they’re arriving to sessions with unusually polished language around their symptoms, terminology they didn’t have three months ago. Or, more starkly, a patient admits they almost cancelled their appointment because the chatbot felt like “enough.” You’re not imagining this shift. It’s happening in practices across the country, and a major new survey finally puts hard numbers to what clinicians have been noticing in their offices.

On June 16, 2026, the American Psychological Association released findings from its 2026 Chatbots and Mental Health Survey, conducted among 1,242 licensed psychologists in April of this year. The results are striking. Seventy-seven percent of psychologists said their patients have used AI for mental health support, companionship, or related purposes. Thirty-nine percent reported patients who had used AI to self-diagnose a mental health condition. This isn’t a fringe phenomenon anymore. It’s sitting in your waiting room.

What Psychologists Are Actually Seeing

The survey paints a picture that goes well beyond someone Googling their symptoms. These are patients forming real patterns of engagement with AI tools, and in some cases, those patterns are creating clinical complications that nobody has a clean playbook for yet.

Thirty-six percent of psychologists in the APA survey reported patients who appeared dependent on a chatbot. That’s more than one in three. Fifteen percent reported patients who developed delusional beliefs after talking to AI. And 13% said patients had formed what they described as an “intimate or relationship-like” connection with an AI chatbot.

Here’s what I tell people when those numbers land hard: the issue isn’t that AI is uniquely evil, or that patients who use it are doing something wrong. The issue is that these tools are being used in ways they were never designed to handle, and patients often don’t know where the line is.

What psychologists are seeing in patients using AI for mental health
Used AI for support77%
Self-diagnosed via AI39%
Appeared chatbot-dependent36%
Developed delusional beliefs15%
Formed intimate AI connection13%
Source: APA 2026 Chatbots and Mental Health Survey

Ninety-four percent of psychologists surveyed said today’s chatbots cannot treat mental health conditions with sufficient nuance, and 89% said chatbots may inadvertently encourage self-harm. Those figures aren’t coming from people who are anti-technology. They’re coming from licensed clinicians watching outcomes in real time.

The Gap Between Using AI and Getting Help

Helpful resource: The Body Keeps the Score by Bessel van der Kolk is a top-rated option for this. (As an Amazon Associate this site earns from qualifying purchases.)

You might be wondering: if people are using AI to talk about their mental health, isn’t that still better than saying nothing at all? It’s a fair question, and the honest answer is: sometimes yes, sometimes no, and the “no” cases are more concerning than we first assumed.

A 2026 KFF survey found that 58% of people who asked AI about a mental health concern did not follow up with a human provider afterward, compared to 42% for physical health questions. That gap matters. For a physical symptom, someone might search online, feel reassured, and reasonably wait it out. Mental health care doesn’t work the same way. The relief someone feels after venting to a chatbot can be real enough to reduce urgency, without addressing anything underneath.

Concern type% who did NOT follow up with a human provider after using AI
Mental health concern58%
Physical health concern42%

This is the substitution problem, and it’s the one that worries clinicians most. AI can feel responsive, available at 2am, non-judgmental, and endlessly patient. Those qualities aren’t nothing. But 58% of people walking away from a mental health concern without human follow-up suggests that for many people, the chatbot conversation is functioning as a destination rather than a starting point.

What This Means If You’re Currently in Therapy

If you’re already working with a therapist, you might be using AI in ways you haven’t mentioned. No judgment here. Lots of people do it. They might process a hard session afterward by typing thoughts into ChatGPT, or use a mental health app between appointments, or ask an AI to help them make sense of a diagnosis they received.

Here’s what I tell people in this situation: bring it up. Not because your therapist will disapprove, but because it’s information that’s relevant to your care. The APA’s June 2026 findings show that 77% of psychologists are already fielding this. Your therapist has almost certainly had this conversation before. If you’ve been using AI to process emotions, work through memories, or explain your feelings in ways that feel easier than saying them out loud, that itself is worth exploring. The pattern you’ve developed with an AI tool can tell a good therapist a lot.

The APA Monitor’s March 2026 piece on AI in the therapist’s office noted that uptake of AI-adjacent tools is increasing even among cautious clinicians, but so is concern about appropriate use. The field is genuinely grappling with where these tools fit.

What This Means If You Haven’t Started Therapy Yet

This one is harder to say, but worth saying clearly. If you’ve been using an AI chatbot as a way to get support without committing to therapy, I understand why. Therapy has real barriers: cost, time, the vulnerability of showing up in front of a stranger. A chatbot removes all of those. But the KFF data showing a 58% non-follow-up rate for mental health concerns suggests that ease of access to AI support may actually be reducing the chances that people seek care that could genuinely change things for them.

AI can be a useful tool for journaling-style reflection, or for finding words for experiences you haven’t named yet. It cannot assess you accurately. It cannot hold the context of your history across sessions the way a trained clinician can. And as the APA survey makes clear, in some cases it can make things worse, not through malice, but through limitations that aren’t visible to the person using it.

If cost is the barrier, options like Open Path Collective (sessions from $30 to $80), community mental health centers, and sliding-scale private practices exist in most areas. SAMHSA’s National Helpline (1-800-662-4357) can help with referrals and is free. A chatbot can help you draft questions to ask a potential therapist. It cannot be the therapist.

For Therapists: This Is Now a Clinical Conversation

The practical upshot of the APA’s June 2026 findings, as covered in detail at HealthPopuli, is that intake conversations and treatment planning now reasonably include questions about AI use. Not as an interrogation, but as a normal part of understanding how a patient is coping between sessions.

Questions worth building in: Are you using any apps or tools to support your mental health between our sessions? Have you talked to any AI tools about what you’ve been experiencing? How did that feel? What did it tell you? The answers can reveal a lot about how a patient understands their own condition, where they’re getting information, and whether any of it needs to be gently corrected.

The therapists who are handling this best, from what I’m seeing, are the ones treating AI use the way they’d treat any coping behavior: with curiosity rather than alarm, and with a genuine interest in what the patient was looking for.

We’re at an early and genuinely uncertain moment with all of this. The tools will keep improving. The ethical questions won’t resolve themselves automatically. What remains constant is that good mental health care depends on a relationship built on accurate information, real accountability, and human judgment. No chatbot has that yet. The 1,242 psychologists who participated in the APA’s April 2026 survey largely agree.


Sources


This article is for general informational purposes only and does not constitute mental health, medical, or clinical advice. If you are in crisis or experiencing a mental health emergency, please contact the 988 Suicide and Crisis Lifeline (call or text 988) or go to your nearest emergency room. Always consult a licensed mental health professional for care specific to your needs.



Disclosure: As an Amazon Associate, we earn a small commission from qualifying purchases at no extra cost to you. We only recommend products that genuinely support the topics covered in this article.