Picture this: a client walks into your office, sits down, and within the first five minutes says, “Well, my AI told me I probably have borderline personality disorder.” The therapist across from them pauses. The client has spent two weeks in deep conversation with a chatbot, arrived with a diagnosis in hand, and now the entire first session is spent unpacking an AI-generated narrative rather than building the therapeutic relationship from the ground up.

This isn’t a rare edge case anymore. It’s Tuesday. It’s every week. And a landmark survey released in June 2026 has finally put hard numbers to what clinicians have been whispering about in supervision rooms and professional Slack channels for months.

The American Psychological Association polled 1,242 licensed U.S. psychologists between April 9 and 26, 2026, and published the results on June 16, 2026. The findings are striking: 77% of psychologists say their patients have discussed using AI chatbots for emotional support, self-diagnosis, companionship, or other mental health purposes. More than a third, specifically 35%, say patients are treating these tools as an additional mental health professional. This isn’t a fringe behavior. It’s reshaping how therapy begins before the client ever books an appointment.

Key takeaways
  • 77% of U.S. psychologists now have patients who discuss using AI for mental health support (APA, June 2026).
  • 35% of psychologists say patients treat AI chatbots as a secondary mental health provider.
  • 39% report patients using AI to self-diagnose before or during therapy.
  • 36% of psychologists have noticed signs of chatbot dependency in their patients.
  • 15% report patients who developed delusional beliefs after AI interactions.

The Chatbot Has Already Done the Intake

What most people don’t realize is that by the time many clients reach a therapist’s office in 2026, they’ve already been “in therapy” for weeks or months. They’ve processed grief with a chatbot at 2 a.m. They’ve described symptoms and received suggested labels. They’ve built a relational pattern with something that responded to them every single time, without judgment, without a billing cycle, and without ever saying “we need to wrap up.”

I’ve seen this pattern accelerate particularly among people who couldn’t access care quickly, people on waitlists, people in rural areas, people who couldn’t afford weekly sessions. The chatbot filled the gap. That’s not inherently malicious. The problem is that these tools were often framing clinical narratives along the way.

The APA survey found that 39% of psychologists have patients who used AI to self-diagnose. That’s not someone Googling their symptoms. That’s an ongoing, conversational process where the AI responded to emotional cues, reflected language back, and gradually built a story about who this person is and what’s wrong with them. By the time the therapist meets the client, they’re often working against an already-set interpretation.

What the Numbers Actually Show

Recommended The CBT Workbook for Mental Health

Evidence-based CBT exercises by clinical psychologists to transform negative thoughts and manage well-being.

Check Price →

Helpful resource: Get Out of Your Mind and Into Your Life (ACT Workbook) is a top-rated option for this. (As an Amazon Associate this site earns from qualifying purchases.)

The survey data, drawn from a sample of 1,242 psychologists pulled from a pool of over 22,000, is specific enough to be sobering. Consider how the concerning behaviors break down:

Behavior reported by psychologists% of psychologists surveyed
Patients discussing AI use for emotional support77%
Patients using AI to self-diagnose39%
Patients treating AI as an additional mental health provider35%
Psychologists noticing signs of chatbot dependency36%
Patients developing delusional beliefs after AI interactions15%
Psychologists saying AI can’t handle mental health with needed nuance94%

That 15% figure deserves its own moment of attention. Delusional beliefs developing after AI interactions. These aren’t patients who got slightly wrong information. These are cases where the line between the AI’s outputs and the client’s reality became blurred in a clinically significant way. That’s a different category of concern than a bad Google search.

Psychologists reporting AI-related patient behaviors (2026)
Discussed AI use77%
Self-diagnosed via AI39%
Treated AI as provider35%
Chatbot dependency signs36%
Delusional beliefs post-AI15%
Source: APA 2026 Chatbots and Mental Health Survey

Here’s something that moves this conversation beyond clinical hand-wringing. In 2026, OpenAI, Google’s Gemini, and xAI have each faced lawsuits tied to chatbot-related mental health harms, including at least one wrongful death claim. Litigation involving AI and mental health is no longer theoretical. It’s in court documents.

This matters for therapists, not because they’re liable for what a chatbot said before the client walked in the door, but because it signals where professional standards are heading. The Forbes analysis published July 9, 2026 noted that therapists often feel underprepared to address AI use clinically and aren’t sure whether, or how, to document it. That gap between what’s happening in sessions and what professional frameworks currently address is going to close, and clinicians who start developing their own approach now will be better positioned.

What Therapists Can Actually Do Right Now

The APA published guidance alongside the survey, specifically a resource on discussing AI use in therapy, also released June 16, 2026. The core recommendation is direct: ask. Make AI use a standard part of intake and ongoing assessment, the way you’d ask about medication or substance use. Not with alarm, with genuine clinical curiosity.

I’ve found the most useful framing is something like: “Some of my clients have been using AI tools between sessions or before starting therapy. Have you? I’m not judging it, I just want to understand your experience.” That question alone can open up a huge amount of information about the client’s expectations, their self-concept, their attachment patterns, and what story they arrived carrying.

When a client has been using a chatbot heavily, the therapeutic work often includes some gentle deconstruction of the AI-generated narrative, not by dismissing the client’s experience, but by helping them hold it more loosely. The chatbot’s label isn’t the truth. It’s a starting point, and often an imprecise one.

For clients who seem dependent on AI contact between sessions, that itself becomes clinically relevant material. It’s worth asking what need the constant availability is meeting. Anxiety containment? Connection? Certainty? Those answers matter more than the chatbot use itself.

The Clients Who Benefit Most From This Conversation

I want to be careful not to frame all AI use as a problem, because that’s not accurate and it’s not fair to clients. For someone who was isolating completely, a chatbot that prompted them to reflect on their mood was a step toward engagement, not a step away from reality. For someone waiting six months for a psychiatry appointment, having a place to process acute distress had some value.

The real concern is when the tool stops being a bridge and becomes the destination. When a client’s entire social-emotional world is mediated by an AI that never pushes back, never has limits, never reflects the complexity of a real relationship, the clinical picture changes. The 36% of psychologists who have noticed dependency signs are describing something that warrants the same gentle, non-shaming attention as any other coping strategy that’s outlived its usefulness.

The conversation isn’t “should clients use AI or not.” That ship has sailed. The conversation is: what happened in that chat window, what did the client take from it, and how does the therapist work with that material honestly?

The clients who benefit most from this conversation in session are exactly the ones who expect you to be incurious about it, who figured the therapist would either dismiss it or be threatened by it. Asking with genuine interest, and sitting with whatever they share, is already therapeutic.

The field is catching up. Slowly, but it’s moving. And in the meantime, the best thing a therapist can do is exactly what they were trained to do: meet the person in front of them, including the parts that arrived with a chatbot’s fingerprints on them.

Sources

Photo: Vitaly Gariev via Pexels

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.