AI Mental Health Statistics 2026: Usage, Research, Access and Privacy

Key AI mental health statistics

How many people use AI for mental health?

Estimates depend heavily on how the question is worded. “Emotional support,” “mental health information” and “a dedicated AI mental health tool” produce very different numbers, so each statistic below keeps the survey’s own wording.

Adults

• 10% of US adults say they ever use AI chatbots for emotional support or advice, and 4% use them for companionship (Pew Research Center, 2026).
• 20% of US adults ages 18 to 29 use AI chatbots for emotional support or advice, compared with 2% of adults 65 and older (Pew Research Center, 2026).
• 16% of US adults used AI tools for mental health information or advice in the past year, including 28% of adults ages 18 to 29 (KFF, 2026).
• Uninsured adults are about twice as likely as insured adults to have relied on AI for mental health advice, 30% versus 14% (KFF, 2026).
• Only 4% of US adults say they have ever used a dedicated AI-powered tool for mental health or emotional well-being (YouGov, 2025).
• 48.7% of US adults with a mental health condition who use AI chatbots used them for mental health support in the past year, in a survey of 499 people (Rousmaniere et al., Practice Innovations, 2025).

Teens and young adults

• 72% of US teens ages 13 to 17 have used AI companions at least once (Common Sense Media, 2025).
• About 1 in 3 teens who use AI companions have discussed serious matters with an AI companion instead of a real person (Common Sense Media, 2025).
• 13.1% of US youths ages 12 to 21, about 5.4 million people, have used generative AI for advice when feeling sad, angry or nervous, rising to 22.2% among those 18 to 21 (McBain et al., JAMA Network Open, 2025).
• 92.7% of young people who used generative AI for mental health advice found it helpful, and 65.5% used it at least monthly (McBain et al., JAMA Network Open, 2025).

Inside general AI chatbots

• 2.9% of Claude.ai conversations are affective, meaning driven by emotional or psychological needs such as counseling, coaching or companionship (Anthropic, 2025).

Why people use AI for mental health support

• 65% of people who use AI for health information say wanting quick or immediate support is a major reason (KFF, 2026).
• 19% of AI health users say not being able to afford a provider is a major reason, rising to 29% of users under 30 (KFF, 2026).
• 90% of people using AI chatbots for mental health support cite accessibility and 70% cite affordability as primary motivations (Rousmaniere et al., 2025).
• 35% of adults ages 18 to 45 who use AI chatbots for mental health say fear of judgment or stigma is their main reason for choosing AI over a professional, ahead of affordability at 32%, in a survey of 400 people (Cognitive FX, 2026).

When people reach out: original data from Abby

In an analysis of anonymized, aggregated Abby conversations from September 22 to October 20, 2026, measured in each person’s local time:

• 1 in 4 messages to Abby (25%) is sent between 10pm and 6am.
• 44% of people have talked to Abby between 10pm and 6am, and 23% between midnight and 4am.
• 72% of messages are sent outside weekday business hours (9am to 5pm), when most support services are closed.
• The busiest hour is 9pm, and Monday is the busiest night of the week.
• People in 126 countries talked to Abby, and 40% live outside the US.
• The typical conversation runs 21 messages, and 73% go past 10 messages.
• Asked what kind of support they want, 61% choose a therapist-style approach and 39% a supportive friend. About 1 in 3 choose more than one.
• 55% of people ask Abby to check in on them first.

Since launch, people have sent Abby more than 10 million messages (lifetime total, October 2026).

Abby is an AI companion for the everyday weight, available at 2am, with free matching to a licensed therapist when you want one. Your first conversation is free.

Start talking to Abby

AI therapy research: what the studies show

The research is promising but early. Results differ between older rule-based chatbots and newer generative AI, and most trials are short.

• In the first randomized controlled trial of a fully generative AI therapy chatbot, 210 adults used Dartmouth’s Therabot or joined a waitlist for 4 weeks. Therabot users had significantly greater reductions in depression, anxiety and eating disorder risk, and rated their therapeutic alliance as comparable to human therapists (Heinz et al., NEJM AI, 2025).
• Dartmouth reported a 51% average reduction in depression symptoms, 31% in anxiety symptoms and 19% in body image and weight concerns among Therabot participants (Dartmouth College, 2025).
• A meta-analysis of 15 randomized trials with 1,744 participants found AI conversational agents reduced depression (effect size g=0.64) and psychological distress (g=0.70), with moderate-quality evidence (Li et al., npj Digital Medicine, 2023).
• A 2025 meta-analysis of 14 randomized trials of generative or hybrid AI mental health chatbots (N=6,314) found a small but significant effect on negative mental health outcomes (0.30) and a larger effect on depression (0.49) (Zhang et al., Journal of Medical Internet Research, 2025).
• A separate 2025 meta-analysis found rule-based chatbots had a small significant effect on depression (g=0.27), while the effect for LLM-based chatbots was not yet statistically significant (Du et al., Journal of Medical Internet Research, 2025).
• In a 2017 randomized trial of 70 young adults, two weeks of CBT-based conversations with the Woebot chatbot significantly reduced depression symptoms compared with an information-only control (Fitzpatrick et al., JMIR Mental Health, 2017).
• In a Stanford-led study, large language models and commercial therapy bots responded inappropriately to mental health scenarios 20% or more of the time, while 16 human therapists responded appropriately 93% of the time (Moore et al., ACM FAccT, 2025).

The mental health access gap

Worldwide

• More than 1 billion people, about 14% of the global population, were living with a mental disorder in 2021 (World Health Organization, 2025).
• Countries have a median of 13.5 mental health workers per 100,000 people, from about 1 in low-income countries to more than 60 in high-income countries (World Health Organization, 2025).
• Two-thirds of countries responding to WHO’s Mental Health Atlas 2024 have one psychiatrist for every 200,000 or more people (World Health Organization, 2025).

United States

• 23.4% of US adults, or 61.5 million people, had a mental illness in 2024 (SAMHSA, 2025).
• Only 52.1% of US adults with a mental illness received any treatment in 2024, leaving 29.5 million untreated (SAMHSA, 2025).
• 154.6 million Americans live in a federally designated mental health professional shortage area, and 7,796 more practitioners would be needed to remove those designations (HRSA, September 2026).
• 46% of US psychologists have no openings for new patients, and 40% keep a waitlist (American Psychological Association, 2025).
• 39% of US psychologists say new clients wait a month or longer for a first appointment, and 13% say four months or more (American Psychological Association, 2025).
• When callers posing as Medicaid patients tried to book psychiatric care in four large US cities, only 27.2% of offices offered an appointment, with median waits from 11 days in Phoenix to 64 days in Los Angeles (JAMA, 2024).
• The average cash price of a therapy session among 175,083 US private-practice therapists is $143.26, and about one-third accept no insurance (Health Affairs Scholar, 2024).

Mental health app privacy statistics

• 19 of 32 popular mental health apps (59%) received Mozilla’s *Privacy Not Included warning label in 2023, down from 28 of 32 in 2022 (Mozilla Foundation, 2023).
• The FTC ordered BetterHelp to pay $7.8 million for sharing users’ health questionnaire answers and other data with advertising platforms after promising to keep it private (Federal Trade Commission, 2023).
• The FTC alleged that Cerebral shared sensitive data on nearly 3.2 million consumers with third parties through tracking tools, and the company agreed to pay more than $7 million (Federal Trade Commission, 2024).
• In a 2026 analysis of 25 popular Android therapy and life-coaching apps, every app contained at least one tracker its privacy policy did not name, and 48% disclosed sending user data to third-party AI providers (Georgiou et al., preprint, 2026).

AI mental health regulation in 2026

• In November 2025 the American Psychological Association advised that generative AI chatbots and wellness apps should not replace a qualified mental health provider, citing a lack of large, high-quality clinical trials (APA, 2025).
• On November 6, 2025, the FDA’s Digital Health Advisory Committee met to advise the agency on generative AI-enabled digital mental health medical devices (FDA, 2025).
• Illinois law, effective August 1, 2025, bars offering therapy through AI unless a licensed professional conducts it, with civil penalties up to $10,000 per violation (Illinois Public Act 104-0054).
• Nevada AB 406 (2025) prohibits AI systems programmed to provide professional mental or behavioral health care, with civil penalties up to $15,000 per violation (Nevada Legislature, 2025).
• Utah regulates rather than bans mental health chatbots: since May 2025 they must disclose they are AI, may not sell user health data, and face fines up to $2,500 per violation (Utah Code 13-72a).
• Maine enacted a law in April 2026 barring therapy through internet-based AI unless a licensed professional provides it (Maine Legislature, 2026).
• Wysa’s 2022 FDA Breakthrough Device Designation covers adults with chronic musculoskeletal pain plus depression and anxiety. It is an expedited review designation, not an FDA approval (Wysa, 2022).
• Woebot Health retired its consumer Woebot app on June 30, 2025 (Woebot Health).

How we compiled these statistics

Every statistic on this page was checked against its primary source: the survey publisher, the peer-reviewed journal, the government agency or the legislature. Where a number appears only in a press release or a non-peer-reviewed preprint, we say so. We left out widely repeated figures we could not trace to a primary source, such as a single “national average wait time” for therapy, and outdated ones, such as “1 in 5 adults” (the 2024 US figure is 23.4%).

Abby’s own statistics come from anonymized, aggregated product analytics from September 22 to October 20, 2026. Times use each person’s profile time zone. Support preferences come from optional onboarding choices. The lifetime message total counts every message people have sent to Abby since launch, through October 2026. Apart from that total, we report percentages and medians only, and no individual conversation was read to produce any of these figures.

We review this page quarterly. Last checked October 21, 2026. You are welcome to cite any statistic here. Please link to this page and to the original source.

Common questions

Abby is not a licensed therapist, psychologist, or psychiatrist, and is not a substitute for professional care. If you are in crisis, call or text 988 (Suicide & Crisis Lifeline) or call 911.