AI in Mental Health Market

In April 2024 the US Food and Drug Administration cleared Rejoyn, the first prescription digital therapeutic authorised for major depressive disorder, and in doing so it moved mental health software out of the wellness app store and into the prescription pad. Douglas Insights values the AI in mental health market at USD 843.2 million in 2025 and expects USD 7.01 billion by 2035, compounding at 23.58%. The estimate is built from the bottom up: about 6,800 provider organisations, payers and employers paying an average USD 124,000 a year for AI-enabled mental health software, triangulated against health system deployments, payer contracts and vendor disclosures. Paying organisations grow 18.6% a year while spend per organisation adds 4.2%, because buyers start with one use case and widen scope rather than paying more for the same tool. This study sits inside Douglas Insights’ healthcare coverage and follows the published Douglas Insights methodology.

Which rules decide what AI in mental health can do?

Regulation sets the ceiling for AI in mental health before any clinical argument does, because a tool that claims to treat a condition is a medical device and a tool that only organises information is not. The Rejoyn clearance in April 2024 marked the line: a prescription digital therapeutic carrying an indication for major depressive disorder went through device review, while the hundreds of meditation and mood-tracking apps beside it did not. That distinction decides the cost of entry. A device pathway brings clinical evidence, quality systems and post-market surveillance, and it also brings the thing vendors actually want, which is a reimbursement code. Douglas Insights treats regulatory status as the primary segmentation driver in this market rather than a footnote, because it separates products that can be prescribed and billed from products sold as employee perks. Buyers checking a specific product’s status should start with the FDA guidance document library, which sets out how software functions are assessed.

What counts as AI in mental health?

AI in mental health covers software that applies machine learning to the detection, treatment, monitoring or administration of mental health conditions, sold to provider organisations, payers, employers and public health bodies. Triage, screening and risk detection covers tools that score symptom severity, flag deterioration and route patients. Digital therapeutics and guided cognitive behavioural therapy covers structured treatment programmes, including prescription products. Clinical documentation and measurement-based care covers ambient scribing, session summarisation and outcome tracking inside the clinical record. Crisis monitoring and population analytics covers helpline triage, suicide risk modelling and the dashboards public health teams use. Value is measured as software and service revenue recognised by vendors. Consumer wellness subscriptions sold direct to individuals sit outside the boundary, as do general-purpose electronic health record systems and pharmaceutical products.

Which regulatory classes and deployment models does AI in mental health cover?

AI in mental health spans three regulatory classes, and cleared medical device software takes about 24% of 2025 value, USD 202 million, Douglas Insights estimates. The classes are cleared medical device software, non-device clinical software, administrative software. Cleared medical device software, including prescription digital therapeutics and some triage tools, carries a regulatory indication and can be billed. Non-device clinical software, about 51% of value, supports clinicians without claiming to diagnose or treat, and administrative software, about 25%, covers clinical documentation, scheduling and reporting. Deployment follows three models: embedded in the electronic health record, standalone clinical platform, patient-facing programme. Tools embedded in the electronic health record take about 38% of value because clinicians use them inside the chart they already open, standalone clinical platforms about 42%, and patient-facing programmes, including guided CBT delivered by app, about 20%. The category list is triage, screening and risk detection, digital therapeutics and guided CBT, clinical documentation and measurement-based care, crisis monitoring and population analytics, and buyers are provider organisations, payers and insurers, employers and education systems, public health bodies and helplines.

What is driving AI in mental health adoption?

Four forces carry the 23.58% revenue path, and the strongest of them is a workforce gap that no hiring plan closes.

The first driver is clinician scarcity. The World Health Organization’s mental health data show how thin provision is outside high-income countries, where median spending sits at a small fraction of health budgets and specialist staffing is measured in single-digit workers per 100,000 people. Douglas Insights counts roughly 6,800 paying organisations in 2025 against a far larger population of eligible providers, which is why the organisation leg grows 18.6% a year: the constraint is triage capacity, and software is the only input available at scale. Waiting lists make the point in numbers: Douglas Insights estimates a typical publicly funded service in Europe or North America carries a 10- to 16-week wait for a first specialist assessment, and a triage tool that re-scores the list weekly lets a service see its highest-risk 5% to 10% of patients first without adding a single clinician.

The second driver is measurement-based care. Payers increasingly pay for outcomes rather than sessions, and outcome payment needs instruments administered on a schedule and scored consistently. That converts a clinical preference into a software requirement, and it explains why documentation and measurement tools hold 22% of 2025 value despite being the least glamorous category in the market.

The third driver is the prescription pathway opened in April 2024. A cleared indication lets a vendor sell into pharmacy benefit and medical benefit budgets rather than the wellness budget, which is both larger and renewed annually. Douglas Insights models digital therapeutics and guided programmes at USD 236.1 million of 2025 value, growing faster than the market average as more products follow the same route.

The fourth driver is ambient documentation. Clinicians spend a substantial share of each session writing notes, and session summarisation tools now cut that materially, which is the rare mental health AI purchase that pays back on administrative time rather than clinical outcome. That makes it the easiest first purchase, and first purchases are what turn a pilot into a paying organisation. Douglas Insights estimates documentation tools are the first purchase in about 45% of new accounts.

What could slow AI in mental health adoption?

Three restraints are modelled. Evidence quality is the first: most products in this market have fewer than two randomised trials behind them, and health systems that were burned by unvalidated pilots now ask for effect sizes before signing, which stretches sales cycles past twelve months. Liability is the second: a triage tool that misses deterioration carries consequences no productivity tool does, so procurement runs through clinical governance and legal review, and vendors carry the insurance cost. Reimbursement breadth is the third: a handful of cleared products can be billed, while the majority are still bought from discretionary budgets that disappear in a cost-cutting cycle, and Douglas Insights holds spend per organisation growth to 4.2% a year for exactly that reason.

Which AI mental health categories carry the value?

Triage, screening and risk detection leads the AI in mental health market with 34% of 2025 value, USD 286.7 million, because every health system with a waiting list needs to know who on it is deteriorating. Digital therapeutics and guided cognitive behavioural therapy hold 28%, USD 236.1 million, a share now underpinned by prescription clearances rather than wellness budgets. Clinical documentation and measurement-based care accounts for 22% of 2025 value, USD 185.5 million, bought first in most organisations because it pays back on clinician time. Crisis monitoring and population analytics is worth USD 134.9 million in 2025, 16% of value, sold mainly to public health bodies and helplines rather than to individual providers.

Which category grows fastest?

Digital therapeutics and guided cognitive behavioural therapy grow fastest within AI in mental health, at an estimated 27.4% a year against a market average of 23.58%, and the reason is billing rather than technology. A cleared indication moves the product from a discretionary wellness line into a reimbursable benefit, and reimbursable products renew. Triage and risk detection grow close behind on health-system demand, while crisis monitoring grows slowest because its buyers are public bodies on fixed budgets and procurement cycles measured in years.

Who buys AI mental health software?

Provider organisations, from hospital behavioural health units to community clinics, buy roughly 54% of 2025 AI in mental health value, and they buy documentation and triage first. Payers and insurers take about 24%, concentrated in digital therapeutics and population analytics where they carry the cost of untreated illness directly. Employers and education systems account for 14%, buying screening and guided programmes as benefits. Public health bodies and helplines take the remaining 8%, small in value but decisive in credibility, because a national helpline deployment is the reference every other buyer asks about.

Where is AI in mental health deployed?

North America leads the AI in mental health market with 48% of 2025 value, USD 404.8 million, growing 21.42% a year, because it combines the deepest behavioural health spending, the clearance pathway that created prescription digital therapeutics, and employers who buy mental health benefits directly. Europe holds 24%, USD 202.4 million, at 24.0%, where national health systems buy centrally and slowly but at scale once convinced, and where medical device rules apply to the same software. Asia Pacific holds 20%, USD 168.6 million, and grows fastest at 27.0%, driven by acute clinician shortages in China and India, high smartphone penetration and public programmes that treat digital delivery as the only affordable route. Latin America contributes USD 33.7 million at 24.5%, the Middle East USD 21.9 million at 25.5% on national wellbeing programmes, and Africa USD 11.8 million at 23.0% from the smallest base.

Who supplies AI in mental health?

Supply splits between specialists and platform vendors. Otsuka and Click Therapeutics developed Rejoyn, the cleared prescription product for major depressive disorder. Big Health sells guided programmes for insomnia and anxiety through payer channels. Woebot Health and Wysa built conversational triage and support tools used by health systems and employers. Spring Health and Lyra Health sell employer-facing platforms with AI triage underneath. Eleos Health applies ambient documentation to behavioural health sessions specifically. Limbic supplies triage into public health services. Douglas Insights estimates the top three vendors hold roughly 21% of 2025 value, a low concentration that reflects how young the category is, and the advantage that lasts belongs to vendors with published trial evidence and a live payer contract rather than to those with the largest user counts. The outside AI vendors that recruit patients and review data for those trials are sized in the AI Based Clinical Trials Solution Provider Market report. The trials that build that evidence run on the platforms sized in our Clinical Trials Management System (CTMS) Market report, and the inpatient alerting that triage tools increasingly feed into is covered in our Hospital Nurse Call Systems Market report.

How is AI in mental health priced?

Average realised spend is USD 124,000 per paying organisation in 2025, and the pricing models behind it differ by category. Documentation tools are priced per clinician, typically USD 1,800 to USD 4,200 a year each. Triage and screening are priced per covered life, from USD 0.40 to USD 2.10 a year, which is why a single large payer contract can move a vendor’s revenue by more than a hundred clinic deals. Prescription digital therapeutics are priced per patient course, in the USD 400 to USD 1,200 band, and are billed rather than expensed. Population analytics is sold as an annual licence, commonly USD 90,000 to USD 350,000. Douglas Insights holds spend growth at 4.2% a year, reflecting scope expansion within accounts rather than list price rises.

Douglas Exclusive: the evidence and reimbursement tracker

The evidence and reimbursement tracker records, for each named product, its regulatory status, the number and size of published trials behind it, whether a billing route exists, and which payers have contracted. Buyers use it to separate products that can be prescribed and reimbursed from products that can only be expensed, which is the distinction that decides renewal.

Scenarios to 2035

The base case pairs 18.6% growth in paying organisations with 4.2% growth in spend per organisation for a 23.58% revenue rate and USD 7.01 billion in 2035. An evidence-gap scenario, in which trials disappoint and health systems retrench to documentation tools only, sets the legs at 11.2% and 1.8%, landing the market near USD 3.05 billion. A reimbursement scenario, in which more products win clearance and payers add codes, sets them at 24.0% and 6.4%, carrying the market past USD 12.4 billion. Each percentage point of organisation growth moves the 2035 figure by roughly USD 640 million. Published estimates for this market cluster between 16% and 34%, a spread that reflects how differently firms draw the boundary around consumer wellness; the Douglas Insights figure sits mid-range because this study excludes direct-to-consumer subscriptions entirely.

Methodology and receipts

The model is built from paying organisation counts by buyer type, realised spend per organisation by category, and deployment evidence from health system and payer announcements, reconciled against vendor revenue disclosures and funding filings. The build covers 41 countries with identifiable deployments, 4 buyer types and 9 vendor disclosure sets, and ties back to about 6,800 organisations at an average USD 124,000 in 2025. Direct-to-consumer wellness subscriptions, general electronic health record systems and pharmaceutical products are excluded. Every figure in the fact sheet carries its own confidence grade, and the working Excel model ships with the licence.

Sources

1. U.S. Food and Drug Administration FDA guidance documents (software functions and digital health) (2026)
2. World Health Organization Global Health Observatory: mental health data (2026)
3. World Health Organization Global Health Observatory: health workforce (2026)