AI Therapy Statistics in US 2026 | Chatbot Usage, Adoption & Facts

What is AI Therapy?

AI therapy refers to the use of artificial intelligence chatbots, apps, and generative AI tools like ChatGPT, Character.AI, and purpose-built platforms to seek emotional support, coping guidance, or mental health information, in place of or alongside a licensed human therapist. In 2026, this is no longer a niche behavior. According to a national KFF Tracking Poll conducted in early 2026, 32% of US adults now say they’ve turned to an AI chatbot for health information in the past year, and 16% specifically for mental health information or advice — a figure that rises to 28% among adults under 30. What began as a curiosity a few years ago has become a genuine, measurable channel through which Americans process anxiety, loneliness, and emotional distress, often outside the traditional clinical system entirely.

The reason AI therapy in America 2026 has grown so quickly is not mysterious. 137 million Americans — 40% of the population — live in a federally designated Mental Health Professional Shortage Area, according to the Health Resources and Services Administration (HRSA), and only 27.29% of mental health workforce need is currently being met in those areas. AI chatbots are free or nearly free, available at 3 a.m., and carry none of the scheduling friction of traditional care. But 2026 has also become the year the safety conversation caught up to the adoption curve: the Federal Trade Commission opened a formal inquiry into seven major AI chatbot companies over child-safety practices, and Illinois, Nevada, and Utah became the first states to restrict or ban AI systems from delivering clinical therapy. This article breaks down exactly what the verified 2026 data shows across usage, safety, effectiveness, market size, and regulation.

Interesting Facts & Latest AI Therapy Statistics in the US 2026

AI THERAPY IN THE US — SNAPSHOT (2026 VERIFIED DATA)
══════════════════════════════════════════════════════════════
  Adults using AI for mental health info   ████████████░░░░░░  16% (32% any health)
  Adults under 30 doing the same           ████████████████░░  28%
  Americans in MH shortage areas (HRSA)    ████████████████░░  137M (40%)
  Psychologists: patients use AI as MH pro ██████████░░░░░░░░  33%
  Teens/young adults (12-21) using AI      █████░░░░░░░░░░░░░  ~13% (up to 22% at 18-21)
  Users satisfied w/ AI mental health resp ████████████████░░  85% (27% "very")
  Global AI mental health market (2026)    ██░░░░░░░░░░░░░░░░  $2.1B → $9.1B by 2033
══════════════════════════════════════════════════════════════

Source: KFF Tracking Poll on Health Information and Trust (Feb–Mar 2026); HRSA Designated HPSA Quarterly Summary (data.hrsa.gov, Q1 FY2026); American Psychological Association 2026 Chatbots and Mental Health Survey; RAND Corporation / JAMA Pediatrics (2026); Grand View Research AI in Mental Health Market Report (2026); Federal Trade Commission press release (Sept 2025)

Two figures in this table define the entire AI therapy story in 2026. The first is the gap between supply and demand: with 137 million Americans sitting in a federally recognized shortage area and only about a quarter of that need actually being met, it is unsurprising that 16% of adults have already turned to a chatbot for mental health information, and nearly 1 in 3 adults under 30 have done the same. The second is the trust paradox: 85% of users report being satisfied with what the AI told them, yet 94% of licensed psychologists say these tools cannot handle the nuance of an actual clinical condition, and 77% of the public remains worried about what happens to the sensitive information they type into these systems. That contradiction — high user satisfaction sitting alongside high professional skepticism and high privacy anxiety — is the defining tension of AI therapy statistics in the US 2026, and it shapes every section that follows.

1. AI Chatbot Usage for Mental Health in the US 2026

WHO USES AI FOR MENTAL HEALTH ADVICE — US ADULTS (KFF, 2026)
══════════════════════════════════════════════════════════════
  All US adults                    ████████░░░░░░░░░░░░  16%
  Adults under 30                  ██████████████░░░░░░  28%
  Adults 50 and older              ████░░░░░░░░░░░░░░░░  8%
  Uninsured adults                 ██████████████░░░░░░  Higher than average
  Black adults                     ████████████░░░░░░░░  Higher than average
  Hispanic adults                  ████████████░░░░░░░░  Higher than average
══════════════════════════════════════════════════════════════

Source: KFF Tracking Poll on Health Information and Trust, conducted February 24–March 2, 2026 (n=1,343 US adults)

The KFF national polling makes clear that AI has quietly reached parity with social media as a source Americans turn to for health guidance, and mental health is a meaningful share of that shift. The 3.5-times gap between adults under 30 (28%) and adults 50 and older (8%) reflects both generational comfort with AI tools generally and, more pointedly, the reality that younger Americans face the steepest barriers to affordable, timely traditional care. The finding that Black adults, Hispanic adults, and uninsured adults are disproportionately more likely to use AI for mental health advice is not incidental — it maps directly onto the same populations that have historically faced the widest gaps in insurance coverage, provider availability, and culturally competent care.

The cost and access data embedded in this table tells its own story: 19% of AI health users cite affordability as a major reason for turning to a chatbot instead of a provider, and 18% cite the simple inability to get an appointment or find a regular doctor. Combined with the fact that 13% of all US adults have already uploaded personal medical information into an AI tool despite 77% expressing privacy concerns, the picture that emerges is one of Americans making a pragmatic trade-off — accepting real privacy risk in exchange for immediate, free, judgment-free access to something, even when they aren’t fully confident in what they’re getting back.

2. AI Therapy Market Size & Growth Statistics in the US 2026

GLOBAL AI MENTAL HEALTH MARKET GROWTH (USD BILLIONS)
══════════════════════════════════════════════════════════════
  2026 market value        ██░░░░░░░░░░░░░░░░░░  $2.1B
  2033 projected value     █████████░░░░░░░░░░░  $9.1B
  Software segment share   ████████████████░░░░  75.8% of revenue (2025)
  N. America market share  ████████░░░░░░░░░░░░  Leading regional segment
══════════════════════════════════════════════════════════════

Source: Grand View Research AI in Mental Health Market Report (2026); National Institute of Mental Health (NIH); National Academy of Medicine “Health in the Headlines” panel (Jan 14, 2026); Rousmaniere et al. research (2025)

The jump from a $2.1 billion market in 2026 to a projected $9.1 billion by 2033 signals that AI-driven mental health tools are moving from experimental add-on to a permanent fixture of the digital health economy, with software products accounting for over three-quarters of all market revenue. What’s notable is where the growth is actually concentrated: it is not primarily happening inside licensed, clinically validated therapy platforms, but inside general-purpose and companion-style products. Character.AI’s roughly 20 million monthly users, more than half of them under 24, illustrates that a huge share of “AI therapy” usage in practice is happening on platforms never designed or regulated as mental health tools at all.

The mid-2025 shutdown of Woebot Health’s consumer chatbot — one of the earliest and most clinically studied AI mental health apps — is a telling counterpoint to the rosy market projections. It suggests that building a standalone, safety-compliant AI therapy product that can operate sustainably is proving far harder in practice than building a general chatbot that people happen to use for emotional support. Meanwhile, the finding that nearly half of adults with a diagnosed mental health condition who used a large language model in the past year did so specifically for mental health support confirms that the underlying demand is real and clinically significant, even if the market’s commercial winners are still shaking out.

3. Youth & Teen AI Chatbot Mental Health Use in the US 2026

AI CHATBOT USE FOR MENTAL HEALTH ADVICE — AGES 12-21
══════════════════════════════════════════════════════════════
  Ages 12-21 overall (2025)         █████░░░░░░░░░░░░░░  ~13%
  Ages 18-21 specifically           ████████░░░░░░░░░░░  22%
  Year-over-year growth (12-21)     ███████████████░░░░  +40%+
  Did NOT disclose use to anyone    ██████████████░░░░░  ~65%
  Found the AI's advice helpful     █████████████████░░  93%
══════════════════════════════════════════════════════════════

Source: RAND Corporation / JAMA Pediatrics, “AI Chatbot Use and Disclosure for Mental Health Among US Adolescents and Young Adults” (published online June 1, 2026); RAND American Life Panel and Ipsos KnowledgePanel survey (Feb–Mar 2025, n=1,058)

The RAND and JAMA Pediatrics research, based on a nationally representative survey panel, is arguably the single most important dataset in this entire topic because it captures something the KFF adult poll cannot: the secrecy built into youth AI mental health use. Nearly two-thirds of young people who used a chatbot for mental health advice never told anyone — not a parent, not a friend, not a counselor — even as 93% found the advice genuinely helpful. That combination of high perceived value and near-total non-disclosure is what researchers describe as the central risk of this trend: young people are getting something they find useful, but adults responsible for their wellbeing generally have no visibility into it at all.

The 40%-plus year-over-year growth in youth chatbot use for mental health advice between 2024 and 2025 is happening against the backdrop of a genuine adolescent mental health crisis, with 18.1% of teens aged 12–17 experiencing a major depressive episode in 2024 and 60% of those teens receiving no treatment whatsoever. Researchers at RAND have been explicit that this is not a coincidence: low cost, constant availability, and perceived anonymity make AI chatbots an obvious stopgap for young people who cannot access — or are afraid to seek — traditional care, even though the same researchers caution that engagement with generative AI raises distinct concerns for those with more intensive or acute mental health needs.

4. AI Therapy Safety, Risk & Regulation Statistics in the US 2026

AI THERAPY SAFETY & REGULATORY ACTIONS — US 2025-2026
══════════════════════════════════════════════════════════════
  AI models responding inappropriately to MH symptoms   ████████████████████  20%+ of tests
  Psychologists: chatbots lack clinical nuance           ██████████████████░  94%
  States banning/restricting AI-delivered therapy        ███░░░░░░░░░░░░░░░░  3 (IL, NV, UT)
  States w/ additional bills advancing (early 2026)      ██████░░░░░░░░░░░░░  6+
  FTC inquiry companies ordered to report                ███████░░░░░░░░░░░░  7 companies
══════════════════════════════════════════════════════════════

Source: Federal Trade Commission press release, “FTC Launches Inquiry into AI Chatbots Acting as Companions” (ftc.gov, September 11, 2025); Moore et al., Stanford University / ACM FAccT (2025); American Psychological Association 2026 Chatbots and Mental Health Survey; state legislative reporting (KFF Health News; Pluribus News)

The Federal Trade Commission’s Section 6(b) inquiry — a formal, compulsory information-gathering action voted through unanimously — is the clearest signal yet that federal regulators view unsupervised AI companion chatbots as a genuine child-safety issue rather than a hypothetical one. The orders required seven major technology companies to detail exactly how they measure, monitor, and mitigate harm to minors, following a string of lawsuits alleging chatbots played a role in cases of self-harm and exploitation among young users. That regulatory response did not emerge in a vacuum: independent Stanford research found that popular AI models gave inappropriate responses to mental health symptoms in at least 1 in 5 tested interactions, and showed measurably more stigmatizing responses toward conditions like alcohol dependence and schizophrenia compared to depression.

At the state level, Illinois, Nevada, and Utah moving first to restrict or outright ban AI systems from delivering clinical therapy — while still permitting AI for administrative and support functions — reflects a legislative consensus that is forming faster than federal policy typically moves. With six or more additional states advancing similar bills as of early 2026, and California, New York, and Texas already imposing disclosure requirements, the regulatory patchwork facing AI mental health products is becoming genuinely complex for any company operating nationally. The near-unanimous 94% of psychologists who say chatbots cannot deliver care with appropriate clinical nuance provides the professional consensus underpinning nearly all of this legislative activity.

5. AI Therapy Effectiveness & Clinical Research Statistics in the US 2026

CLINICAL RESEARCH ON AI THERAPY TOOLS — US 2025-2026
══════════════════════════════════════════════════════════════
  Therabot RCT: depression symptom reduction    ████████████████████  51% average
  AI users who also attend traditional therapy  ██████████████░░░░░░  71%
  Adults w/ MH condition using LLM for support  ██████████████████░░  48.7%
  Psychologists: patients use AI as add'l pro   █████████████░░░░░░░  33%
  Americans hopeful about AI in mental health   █████████░░░░░░░░░░░  31%
══════════════════════════════════════════════════════════════

Source: Heinz et al., “Therabot” randomized controlled trial, NEJM AI (2025); American Psychological Association 2026 Chatbots and Mental Health Survey; YouGov survey (April–May 2026); Rousmaniere et al. (2025)

The Therabot trial stands apart from nearly every other data point in this article because it is the first randomized controlled trial of a purpose-built generative AI therapy chatbot, meaning it meets the same evidentiary bar used to evaluate human-delivered treatments. A 51% average reduction in depression symptoms is a clinically meaningful result, and it is the single strongest piece of evidence that a properly designed, narrowly scoped AI therapy tool can work. But that finding should not be generalized to the broader category of chatbots people are actually using — the same evidence base that supports Therabot is precisely what’s missing from general-purpose tools like ChatGPT or Character.AI, which is why the 71% of AI therapy users who also attend traditional therapy is the more representative real-world pattern: AI as a supplement, not a substitute.

Public sentiment reflects this uncertainty almost perfectly. Americans split nearly evenly three ways — hopeful, anxious, and neutral — about AI’s role in mental health care, and concern is actually rising among both younger and middle-aged adults according to YouGov polling through 2026. That rising anxiety, even as usage climbs, suggests Americans are adopting these tools out of necessity and convenience rather than confidence, a dynamic reinforced by cited MIT research warning that relying on AI alone for emotional support may, for some users, deepen rather than relieve loneliness.

6. AI Therapy Demographics, Trust & Adoption Trends in the US 2026

WHO USES AI FOR MENTAL HEALTH & HOW MUCH THEY TRUST IT — 2026
══════════════════════════════════════════════════════════════
  Users trusting AI "a great deal/fair amount" (MH)  ████████████░░░░░░░░  62%
  Users who DON'T trust AI for reliable MH info      ███████████░░░░░░░░░  38%
  Non-users who distrust AI for health info entirely ████████████████░░░░  ~80%
  Users following up with a provider after asking AI ████████████░░░░░░░░  42% (physical) vs lower for MH
  Adults using AI several times a week (any purpose  ████████████████░░░░  39%
══════════════════════════════════════════════════════════════

Source: KFF Tracking Poll on Health Information and Trust (Feb–Mar 2026); EBRI Employer Mental Health Survey (2025)

The trust data here reveals a self-reinforcing loop that shapes almost every other statistic in this article: people who have never used AI for mental health support are deeply skeptical of it (roughly 80% distrust), while people who have actually used it report meaningfully higher trust (62%) — even though a sizable minority of users, 38%, remain unconvinced even after trying it. This gap between users and non-users suggests that AI mental health tools benefit from a kind of self-selecting adoption curve: the people trying them tend to already be somewhat more open to the format, which can make aggregate satisfaction numbers look stronger than they’d be if the least AI-receptive Americans were using these tools at the same rate.

Perhaps the most consequential number in this section is the 58% of users who did not follow up with a human provider after asking an AI chatbot about a mental health concern, notably higher than the equivalent figure for physical health questions. For readers tracking the same access crisis documented in our broader therapy statistics in the US coverage — where cost and provider shortages remain the top-cited barriers to care — this non-follow-up rate is exactly the outcome clinicians and regulators worry about most: AI functioning as an endpoint rather than a bridge to appropriate treatment. That risk compounds further when layered against our findings on the broader mental health statistics in the US, where treatment gaps already run into the tens of millions of untreated adults, and against youth-specific patterns detailed in our student mental health statistics coverage, where a majority of depressed teens already receive no formal care at all. Whether AI closes that gap or simply gives people a plausible-feeling reason to delay real treatment is the single biggest open question in AI therapy statistics for the US in 2026 — and one that regulators, researchers, and clinicians will be racing to answer through the rest of this decade.

Disclaimer: This research report is compiled from publicly available sources, including federal data, national surveys, and peer-reviewed research. While reasonable efforts have been made to ensure accuracy, no representation or warranty, express or implied, is given as to the completeness or reliability of the information. We accept no liability for any errors, omissions, losses, or damages of any kind arising from the use of this report.

Email ID:

Please enable JavaScript in your browser to complete this form.

Email

Email *

SUBSCRIBE