Survey finds mental health patients turning to AI chatbots amid care gaps

A national survey of 416 behavioral health clinicians finds 9% to 10% say patients regularly or occasionally disclose using AI for emotional support, with 32% of providers reporting possible or actual harm. The findings point to cost, wait times and access barriers pushing patients toward commercial chatbots instead of licensed care.

Why it matters: - Patients are using commercial AI chatbots for emotional support when licensed care is too costly, too slow or unavailable. - Clinicians say that shift is already causing harm for some patients. - The trend raises questions about whether AI should be used as a substitute for therapy or as a back-end tool to ease clinician workload.

What happened: - ICANotes released the ICANotes Clinician Survey 2026 on July 8, 2026. - The national survey covered 416 licensed behavioral health professionals across the U.S. - 9% to 10% of clinicians said their patients now regularly or occasionally disclose using AI for emotional support. - 18% of clinicians said they have already treated patients whose mental health outcomes worsened with unguided AI use. - 14% reported observing possible clinical harm.

The details: - Providers identified high out-of-pocket therapy costs as a leading reason patients try algorithmic support, at 21%. - Another 21% pointed to patients not recognizing the severity of their need for licensed human care. - 11% cited an inability to reach a human provider during an immediate crisis. - 23% of behavioral health providers reported waitlists of four or more weeks or said they were closed to new patients. - ICANotes Chief Clinical Officer October Boyles said patients are turning to AI because the human healthcare system is failing to meet them in their moment of need. - Boyles said a smartphone app can look appealing during a crisis at 2 a.m. when an appointment may be weeks away. - Frontline clinicians said patients often try AI apps because they are available late at night when no one else is reachable.

Between the lines: - The survey argues the core problem is not that patients prefer AI over human therapists. - The bigger issue is access, with financial barriers and wait times steering people to always-on technology. - The findings suggest mental health tech may be safer and more useful if it reduces administrative burden rather than simulating therapy. - The survey says administrative AI could help cut documentation work that consumes a full workday per week for 40% of clinicians.

What's next: - The findings are aimed at tech developers, health executives and bioethicists gathering at AI conferences this month. - The likely pressure point is whether companies keep building front-end chatbot therapists or shift toward back-end tools that free up clinician time. - The broader outcome will depend on whether health systems can reduce waitlists and expand access before more patients turn to unsupervised AI support.

The bottom line: - The survey frames AI chatbots as a stopgap born from a strained mental health system, not a safe replacement for care.