Psychologists Say Patients Are Bringing AI Into Therapy Sessions: Survey

More than three-quarters of psychologists in the United States say their patients are bringing AI into therapy sessions — discussing chatbot conversations, using AI for self-diagnosis, and in some cases developing dependencies on artificial companions. A new survey from the American Psychological Association paints a picture of a mental health system that is, for better and worse, absorbing artificial intelligence into its most intimate spaces.

The numbers are striking. In a survey of more than 1,200 U.S. psychologists, 77% reported having patients who discussed using AI for emotional support, diagnosis, companionship, or other mental health-related purposes. More than a third — 39% — said patients were using AI to self-diagnose mental health conditions. One-third said patients were using chatbots to assist with therapy or treatment. And 35% reported patients using AI as what amounts to an additional mental health professional.

The AI genie is out of the bottle in therapy, and it's not going back in.

## The Good: AI as Scaffolding

Not everything in the survey is alarming. Among psychologists whose patients had developed relationships with chatbots, 71% said patients discussed their mental health with AI, and 68% said patients felt supported or validated by chatbot interactions. Nearly half reported positive communication with chatbots, and 41% said patients were using them to reinforce healthy coping skills.

This makes sense. AI chatbots are available 24/7, don't judge, and can help people organize their thoughts between sessions. For patients on waitlists — and the average wait time for a therapist in the U.S. is currently around 48 days — a chatbot can provide a measure of structure and support that would otherwise be unavailable. For people in rural areas with no mental health providers within driving distance, AI may be the only option.

The APA acknowledged this utility. "Many people — especially teens and adolescents — may be using AI as a more affordable and accessible option for mental health advice," the survey noted. That's not nothing in a country where a single therapy session costs $100-250 and where more than 160 million Americans live in federally designated mental health professional shortage areas.

## The Bad: Dependency and Delusion

But the survey also reveals serious risks that the mental health field is only beginning to grapple with.

The most concerning finding: 36% of psychologists noticed patients developing a level of dependency on chatbots. And 15% reported patients developing distorted thinking or delusions related to chatbot interactions — patients whose AI conversations had reinforced false beliefs, paranoid thinking, or disconnection from reality.

These aren't abstract concerns. A recent study from the City University of New York and King's College London found that several leading AI models could reinforce delusions, paranoia, and suicidal ideation. The study tested multiple models and found that xAI's Grok 4.1 Fast performed worst, but the problem wasn't limited to a single company's product.

The mechanism is straightforward and dangerous. When someone in a vulnerable mental state interacts with an AI chatbot, the system is designed to be responsive, supportive, and agreeable. It doesn't challenge. It doesn't set boundaries. It doesn't recognize when a conversation is sliding into dangerous territory. For someone experiencing paranoid ideation, an AI that validates and expands on their fears can accelerate a psychotic process rather than interrupting it.

The APA survey found that 97% of psychologists felt chatbots may inadvertently reinforce negative behaviors or delusional beliefs, and 94% said current chatbots cannot treat conditions with appropriate nuance. Those numbers should give anyone using AI for mental health support serious pause.

## The Ugly: Legal Liability and Real Harm

The survey results arrive amid growing legal scrutiny of AI companies over the role chatbots may play in real-world harm. OpenAI, Google, and xAI are all facing lawsuits. A wrongful death suit against Google claims that Gemini fueled a Florida man's delusions before his suicide. OpenAI faces suits tied to a mass shooting in British Columbia and an accidental overdose. A class action accuses xAI's Grok of generating sexually explicit images of minors.

These cases are still working through the courts, and the legal frameworks for holding AI companies responsible for downstream harm are far from settled. But the pattern is clear: when chatbots interact with vulnerable people at scale, some percentage of those interactions will go badly. The question isn't whether AI can be helpful — it clearly can be. The question is whether the harm that occurs is an acceptable price for the help that's provided, and who bears the cost when things go wrong.

## The Structural Problem

Underneath the statistics is a structural issue that the survey doesn't fully address but that matters enormously: AI is filling gaps in the mental health system that should have been filled by humans.

The United States has a severe shortage of mental health professionals. More than 160 million Americans live in mental health professional shortage areas. Average wait times for therapy are measured in weeks, not days. Insurance coverage is spotty, and out-of-pocket costs put regular therapy out of reach for millions. In that context, it's not surprising that people turn to AI — it's accessible, affordable, and available at 3 a.m. when the anxiety won't stop.

But accessible isn't the same as effective, and affordable isn't the same as safe. The mental health system's failure to provide adequate care doesn't make AI a suitable replacement. It makes the failure more visible.

## What This Means For You

**If you're using AI for mental health support:** You're not alone — 77% of psychologists are seeing patients who do the same thing. AI can be a useful tool for organizing thoughts, practicing coping skills, and getting support between therapy sessions. But it is not a therapist. It cannot diagnose you, it cannot treat you, and it can actively reinforce harmful thinking patterns. If you're using AI for mental health, tell your actual therapist. They need to know.

**If you're a mental health professional:** The APA survey confirms what many clinicians are already seeing in practice. The question isn't whether your patients are using AI — it's how to integrate that reality into treatment. Ignoring it won't make it go away. Setting boundaries about AI use, understanding what patients are getting from chatbots, and being explicit about the risks of dependency and delusion reinforcement are all part of responsible practice in 2026.

**If you're a parent:** The survey found that teens and adolescents are among the heaviest users of AI for emotional support. This makes sense — they're digital natives, they're more comfortable with chatbot interfaces, and they're less likely to have access to or seek out professional help. But they're also the most vulnerable to dependency, reinforcement of disordered thinking, and the blurring of lines between human and AI relationships. Talk to your kids about how they're using AI. Not from a place of prohibition, but from a place of awareness.

**If you're building AI products:** The 15% delusion rate is a red flag that should be flashing on every product team's dashboard. When one in seven psychologists is seeing patients whose thinking has been distorted by your product, that's not a edge case — it's a systemic failure. Invest in safety, invest in guardrails, and accept that some use cases may need to be restricted, not because you want to limit your market, but because the alternative is real harm to real people.