AI Is Infiltrating Therapy Rooms—And Psychologists Are Divided on Whether That's Good or Bad

More than a third of psychologists now report that their patients are bringing AI into clinical sessions as a supplementary mental health tool, according to a fresh survey from the American Psychological Association. This trend reveals a widening gap between patient adoption of artificial intelligence and clinical best practices—a friction point we're watching closely in how technology intersects with professional healthcare delivery.

The data signals something significant: patients aren't waiting for their therapists' approval before integrating AI into their treatment regimen. They're treating chatbots and language models as an extension of their care team, consulting them between sessions or even during therapeutic work. For some patients, AI provides accessible, on-demand support when traditional therapy appointments feel sparse or expensive.

But here's where it gets thorny. Leading psychologists are sounding alarms about a critical risk: AI systems can inadvertently reinforce delusions and distorted thinking patterns. Rather than challenging unhealthy thought spirals—a core function of evidence-based therapies like cognitive behavioral therapy—language models might validate and amplify them. A patient with paranoid ideation could feed that concern into an AI, receive a response that doesn't contradict it, and leave the interaction more convinced of their original fear.

The disconnect is real. Patients are seeking efficiency and around-the-clock availability that traditional therapy can't match. Insurance limitations, therapist shortages, and stigma around mental health still keep millions from accessing care at all. AI fills that gap—or appears to. The problem is substitution: when AI becomes a replacement for professional judgment rather than a true supplement, outcomes deteriorate.

What This Means for Treatment Quality

The survey underscores a broader challenge in mental health: patients have agency over their treatment decisions, but not all those decisions land them in the right place. A therapist can contextualize patient concerns, distinguish between valid insights and distortions, and adjust interventions on the fly. AI cannot—at least not with the nuance required for complex psychological work.

Clinicians are watching this adoption curve with justified concern. They're trained to recognize patterns that language models miss. They understand that certain psychiatric conditions—schizophrenia, delusional disorder, severe anxiety—require sophisticated clinical reasoning to avoid harm.