People are turning to AI for mental health support faster than the evidence, the safety guardrails, or the law can keep up. This page collects the most reliable AI in mental health statistics for 2026 — how many people use AI for mental health, what the clinical research shows, the access gap driving demand, and the safety and regulatory backlash — with each number checked against the survey, peer-reviewed study, court ruling, or government source that published it.

The short answer: About 16% of U.S. adults (and 28% of under-30s) have used AI chatbots for mental health information (KFF), and roughly 1 in 8 teens and young adults have used AI for mental health advice (JAMA Network Open). The first randomized trial of a purpose-built therapy chatbot showed real symptom reductions (Dartmouth/NEJM AI) — but high-profile harms have prompted states like Illinois to ban AI from delivering therapy while still permitting AI for administrative support.

If you or someone you know is struggling or in crisis, call or text 988 (the U.S. Suicide & Crisis Lifeline) for free, confidential, 24/7 support. AI tools are not a substitute for professional care.

Top AI in mental health statistics for 2026 (editor's picks)

• 16% of U.S. adults used an AI chatbot for mental health information in the past year — 28% among adults under 30. — KFF, 2026

• 1 in 8 young people (ages 12–21) have used AI for mental health advice; 93% of them found it helpful. — JAMA Network Open, 2025

• ~51% and ~31% reductions in depression and anxiety symptoms in the first RCT of a purpose-built therapy chatbot. — Dartmouth / NEJM AI, 2025

• 48.7% of adults with a mental health condition who used LLMs used them for mental health support. — Rousmaniere et al., 2025

• Nearly half of U.S. adults with any mental illness received no treatment in the past year. — SAMHSA, 2025

• First in the nation: Illinois banned AI from providing therapy in 2025 — while expressly allowing AI for administrative support. — Illinois WOPR Act, 2025

• Only 16% of large-language-model mental-health chatbot studies have undergone clinical efficacy testing. — Systematic review, 2025

• 58% of people who asked AI about a mental health concern did not follow up with a provider. — KFF, 2026

How many people use AI for mental health

• 16% of U.S. adults sought mental health information or advice from an AI chatbot in the past year, part of the 32% who used AI for any health information. — KFF, 2026

• 28% of adults under 30 used AI for mental health information — about three times the rate of adults 50 and older. — KFF, 2026

• Black, Hispanic, and uninsured adults were more likely than others to turn to AI chatbots for mental health advice. — KFF, 2026

• 1 in 8 (13%) of young people aged 12–21 reported using AI for mental health advice; among 18–21-year-olds it rose to 22%. — JAMA Network Open, 2025

• 93% of those young users found the AI's advice helpful, and 66% used it at least monthly. — JAMA Network Open, 2025

• 48.7% of adults with a mental health condition who had used a large language model in the past year used it for mental health support. — Rousmaniere et al., 2025

• 20 million monthly users engage with Character.AI, more than half of them under 24. — reported in APA, 2026

What the clinical research shows

The strongest evidence comes from one landmark trial — and it comes with heavy caveats.

• First-ever RCT: Dartmouth researchers ran the first randomized controlled trial of a generative-AI therapy chatbot ("Therabot"), with 210 adults who had significant depression, anxiety, or eating-disorder risk. — Dartmouth / NEJM AI, 2025

• ~51% reduction in depression symptoms, ~31% in anxiety, and ~19% in eating-disorder concerns versus a waitlist control. — Dartmouth / NEJM AI, 2025

• Comparable alliance: participants rated their therapeutic alliance with the chatbot as similar to a human therapist, and engaged with it for ~6 hours on average. — Dartmouth / NEJM AI, 2025

• Expert-built, not off-the-shelf: Therabot was fine-tuned on CBT and psychotherapy best practices by a clinical team — it is not a general chatbot like ChatGPT, and the researchers stressed clinician supervision is essential. — Dartmouth / NEJM AI, 2025

• A thin evidence base: in a systematic review of 160 chatbot studies, LLM-based chatbots jumped to 45% of new studies in 2024, yet only 16% of LLM studies underwent clinical efficacy testing (77% remained in early validation). — Systematic review, 2025

• Worse as severity rises: across studies comparing chatbots to human therapists for moderate-to-severe anxiety or depression, human therapists produced significantly better outcomes — and the gap widens with severity. — Research review, 2026

The access gap driving demand

AI is filling a vacuum created by a system that can't meet demand.

• 61.5 million U.S. adults (23.4%) experienced any mental illness in the past year; 14.6 million had a serious mental illness. — SAMHSA, 2025

• Nearly half (about 48%) of adults with any mental illness did not receive mental health treatment in the past year. — SAMHSA, 2025

• 80% of people who needed substance-use treatment did not receive it. — SAMHSA, 2025

• Cost and provider availability rank among the top reasons people don't get care, alongside stigma and "thinking they can handle it on their own." — SAMHSA, 2025

• Roughly 40% of the U.S. population lives in a designated Mental Health Professional Shortage Area, where only about a quarter of the need for care is met. — HRSA, 2025

The safety and regulatory backlash

Documented harms have moved this from a research debate to active law and enforcement.

• First-in-the-nation ban: Illinois's Wellness and Oversight for Psychological Resources Act (2025) prohibits using AI to provide therapy or make therapeutic decisions. — Illinois WOPR Act, 2025

• Administrative AI is expressly allowed: the same law permits AI for administrative and supplementary support for licensed behavioral health professionals — drawing a clear line between clinical and operational use. — Illinois WOPR Act, 2025

• A growing trend: Nevada and Utah passed similar restrictions in 2025, with California, New Jersey, and Pennsylvania considering measures. — reported across legal analyses, 2025

• Federal scrutiny: the FTC issued orders to seven major AI chatbot companies in September 2025 seeking information on their safety and data practices. — FTC, 2025

• Lawsuits over minors: the APA has cited lawsuits involving minors who used AI chatbots for "therapy," including a case ending in suicide, warning that AI masquerading as a therapist can put lives at risk. — APA, 2025

• Market exit: Woebot Health shut down its consumer-facing CBT chatbot in mid-2025, a signal of how hard standalone AI "therapy" has become to operate responsibly. — reported, 2025

How people feel about AI for mental health

• 85% of users were at least somewhat satisfied with AI responses to mental health questions, though only 27% were "very satisfied." — KFF, 2026

• 62% of those who used AI for mental health information said they trust it "a great deal" or "a fair amount" — but 38% of users said they don't trust it. — KFF, 2026

• 58% of people who asked AI about a mental health concern did not follow up with a provider afterward (vs. 42% for physical health). — KFF, 2026

• 77% of adults are concerned about the privacy of personal medical information shared with AI tools. — KFF, 2026

Where AI fits responsibly: the operations layer

The law and the evidence point to the same conclusion: AI should not act as a therapist, but it can take the administrative load off the humans who do. Behavioral health practices are buried in phone work — intake calls, appointment scheduling, reminders, insurance questions, and after-hours messages — at the same time clinicians are in critically short supply. That's a non-clinical problem AI is well-suited to.

This is exactly the line Illinois's law draws: AI can't deliver therapy, but it can handle administrative and support functions for licensed providers. A voice agent answering scheduling and intake calls — and routing anything clinical or urgent to a human — keeps front-desk operations running without ever touching care decisions.

Brilo AI voice agents handle the operational calls a practice can't keep up with — booking and confirming appointments, capturing intake details, and routing clinical or sensitive matters to your staff — never providing therapy or clinical advice. See how Brilo AI supports healthcare practices with appointment booking and intake, so clinicians spend their time on patients, not the phone.