More than one billion people are currently living with a mental health condition. The number of professionals available to treat them has not come close to keeping pace. The global median stands at just 13 mental health workers per 100,000 people — a figure the WHO’s September 2025 Mental Health Atlas describes as critically inadequate, with extreme shortages concentrated in low- and middle-income countries. WHO estimates a shortage of approximately 1.18 million mental health workers globally to meet current needs — a figure that does not account for the additional burden created by the COVID-19 pandemic’s lasting effects on population mental health.
The gap between need and supply is not primarily a problem technology is positioned to solve, however confidently the AI industry suggests otherwise. It is a problem of funding, workforce pipeline, insurance architecture, and political prioritisation — and in 2026, all four of those systems are moving in the wrong direction in at least some of the places that matter most.
Further reading: The Global Loneliness Epidemic: Why Governments Are Now Treating Isolation as a Public Health Crisis
The Supply Problem Is Worse Than the Headlines Suggest
The headline shortage figure understates the real picture in two ways. First, the workers who do exist are distributed catastrophically unevenly. Malawi, a country of 20 million people, has just four psychiatrists. The United States, with 330 million people, has tens of thousands, yet 40% of its population lives in federally designated Mental Health Professional Shortage Areas. The shortage is not simply a global arithmetic problem; it is a structural maldistribution problem that leaves communities underserved even in wealthy countries with relatively large workforces.
Second, the pipeline is not keeping up with demand growth. US federal projections indicate a 49% increase in demand for mental health services by 2033, while workforce supply is projected to grow by only 11% over the same period. The projected shortage of 88,000 mental health counselors and 114,000 addiction counselors by 2037 represents a workforce deficit of more than 200,000 practitioners across just two disciplines. Marriage and family therapists face a projected adequacy rate of just 45% — meaning the profession will be able to meet fewer than half of projected demand by the mid-2030s.
The reasons the pipeline is failing are structural rather than incidental. The mental health workforce shortage is a vicious cycle in which burnout, inadequate reimbursement, and protracted licensure processes each worsen the others — and effective solutions must tackle the workforce pipeline, clinician retention, and payment reform simultaneously rather than piecemeal, because each factor reinforces the others. Therapist burnout rates have risen alongside patient caseloads, creating an attrition problem that recruitment alone cannot offset. In the US, the average time to full licensure for a clinical psychologist now exceeds a decade of combined education and supervised training — a barrier that narrows the profession’s intake in ways that compound over the years.
The WHO’s September 2025 reports also found that fewer than 10% of countries have fully transitioned to community-based mental health care models, with most still in early stages, and that nearly half of all psychiatric hospital admissions occur involuntarily — a figure reflecting systems still organised around acute crisis rather than ongoing community support.
The Insurance Failure: A Law on the Books, Barely Enforced
In the United States, where the insurance architecture is the primary determinant of access for most working-age adults, the system designed to guarantee mental health coverage has spent 2025 and 2026 moving backwards.
The Mental Health Parity and Addiction Equity Act — passed in 2008, strengthened by a final rule in September 2024 — requires insurers to cover mental health services on terms no more restrictive than physical health services. The principle is straightforward. The implementation has been a sustained failure. Insurance reimbursements for behavioural health visits are, on average, 22% lower than for equivalent medical or surgical office visits, creating a structural disincentive for providers to join insurance networks that has persisted despite the parity law’s existence for nearly two decades.
The 2024 strengthening rule, designed to address this gap through outcome-based compliance measures, has effectively been suspended. In May 2025, the US Departments of Labor, Health and Human Services, and Treasury announced they would not enforce the final rule before courts resolved an ongoing legal challenge — a challenge brought by ERIC, an employer group that asserted it “should have no obligation to ensure there is no material difference in access to substance use disorder and mental health services.” The Trump administration has since indicated it will propose a new rule no later than December 2026 to replace the suspended 2024 version, leaving the regulatory landscape in limbo for the intervening period.
The practical consequences for patients are documented and severe. One in four insured Americans cannot find a mental health therapist in their health plan’s network. Out-of-pocket costs for outpatient mental health care significantly exceed those for equivalent medical specialty care, and eight in ten respondents to NAMI surveys reported out-of-pocket costs exceeding $200 for psychiatric hospital or residential care — compared to fewer than six in ten for general hospital care.
The most striking recent enforcement action illustrates how far the gap between legal obligation and operational reality extends. In February 2026, Kaiser Foundation Health Plan settled with the US Department of Labor for failing to maintain adequate provider networks for mental health and substance use disorder care — agreeing to pay a $2.8 million penalty and compensate at least $28 million to members who had been forced to seek out-of-network mental health services at their own expense.
Georgia’s insurance commissioner logged more than 6,000 parity violations across 22 insurers in 2025 and 2026 — the largest state-level enforcement action in the parity law’s history, accounting for $25 million of the $31 million in total fines levied by ten states over the past six years. That the most aggressive enforcement in the law’s 17-year history is happening at the state level, in Georgia, driven by a Republican commissioner, while federal enforcement is suspended, captures the current policy landscape with some precision.
LMICs: Where “Shortage” Understates Total Absence
If the US insurance system represents a case where legal protections exist but fail to reach patients, the situation in low- and middle-income countries represents something closer to the near-total absence of any system at all.
WHO estimates that over 75% of those needing mental health treatment in LMICs receive no treatment whatsoever. LMICs allocate a disproportionately small fraction of healthcare budgets to mental health, with a treatment gap exceeding 75%, and median annual government expenditures for mental health in low-income countries estimated at just eight cents per capita. For context, the US spends approximately $280 per capita annually on mental health services, itself considered inadequate by most clinical benchmarks.
Over 87% of all patients experiencing first-episode psychosis live in LMICs, where mental health services are scarce, and the treatment gap remains alarmingly wide — with the average duration of untreated psychosis in LMICs running approximately 48 weeks, compared to 41 weeks in high-income countries. Seven extra weeks of untreated psychosis, on average, for every first-episode patient in a lower-income country. Multiplied across hundreds of millions of people, that gap represents an incalculable burden of preventable suffering.
The dominant response to this gap from global health institutions has been task-shifting — training community health workers, nurses, and primary care physicians to identify and treat common mental disorders using simplified WHO-developed protocols like the Mental Health Gap Action Programme. The evidence for this approach is mixed: task-shifting can increase access to basic care, but it cannot substitute for specialist expertise in complex or severe cases, and it requires sustained training, supervision, and quality assurance investment that most LMIC health systems cannot consistently provide.
What AI Can and Cannot Do for Mental Health
Against this backdrop, the technology industry’s pivot toward AI-powered mental health tools deserves clear-eyed assessment rather than either uncritical enthusiasm or reflexive dismissal.
The evidence base is genuinely emerging. A March 2025 randomised controlled trial of Therabot, a generative AI therapy chatbot developed at Dartmouth, found clinically significant reductions in symptoms among patients diagnosed with Major Depressive Disorder, Generalised Anxiety Disorder, and high-risk feeding and eating disorders. Digital mental health platforms processed approximately 100 million therapy sessions globally in 2025 — a scale that no human workforce could match.APA’s 2026 Chatbots and Mental Health Survey of more than 1,200 licensed psychologists found that more than a third reported patients turning to AI as a supplementary mental health resource, confirming that adoption is already happening regardless of what clinicians or regulators recommend.
The limitations, however, are equally documented. AI chatbots cannot replicate genuine empathy, risk misinterpretation during crises, rely on short-term evidence, and lack standardised evaluation frameworks — with privacy and data security representing significant ethical concerns that remain unresolved. A March 2026 meta-analysis in npj Digital Medicine found that among 39 included chatbot studies, 23 reported no systematic safety monitoring or adverse-event data whatsoever — meaning the field is racing ahead of its own safety infrastructure. In the Therabot trial itself, staff had to intervene 15 times for participant safety concerns, including expressions of suicidal ideation, and 13 additional times to correct inappropriate chatbot responses.
The structural argument against AI-as-fix is straightforward. The mental health crisis in LMICs is not primarily caused by a shortage of therapy sessions at scale — it is caused by a shortage of funding, trained specialists, functioning health systems, and social infrastructure. An AI chatbot that requires a smartphone, reliable internet, digital literacy, and the ability to engage in therapeutic dialogue in a supported language does not reach the populations with the highest burden. And in the United States, where the insurance parity law is suspended, and 137 million people live in designated shortage areas, adding AI tools does not fix the underlying reimbursement and network access failures that determine whether most people can access care at all.
The honest conclusion from the evidence is that AI can function as a complement to human mental health services — expanding access for some populations, providing support between sessions, and potentially reaching people who would not otherwise seek care — but it cannot substitute for the systemic reforms that the workforce shortage actually requires: better reimbursement, shorter licensure pathways, community-based care models, and sustained public investment in a field that has been chronically underfunded across the income spectrum.
The Mental Health Reckoning the System Keeps Postponing
The WHO’s September 2025 Mental Health Atlas was published ahead of the UN High-Level Meeting on noncommunicable diseases and mental health in New York that month. WHO Director-General Dr Tedros Adhanom Ghebreyesus said at the meeting that transforming mental health services is “one of the most pressing public health challenges” and that “investing in mental health means investing in people, communities, and economies — an investment no country can afford to neglect.”
The gap between that framing and the actual policy trajectory — suspended parity enforcement in the US, eight cents per capita in low-income countries, a projected 2037 US workforce that will meet less than half of demand for marriage and family therapy — is not a communications problem. It is a prioritisation problem. Mental health has never attracted the same combination of political attention, insurance investment, and public health infrastructure that cardiovascular disease, cancer, and infectious disease have commanded, despite carrying a comparable share of the global burden of illness.
The billion people living with mental health conditions in 2026 are not waiting for a better chatbot. They are waiting for the systems that were promised to them to actually be built.
Sources: WHO Mental Health Atlas 2024 (published September 2025); Health Resources and Services Administration (HRSA), State of the Behavioral Health Workforce, 2025; Healing Psychiatry of Florida, Mental Health Workforce Shortage Statistics 2026 (May 2026); Commonwealth Fund, “Behavioral Health Parity Takes Step Backward Under Trump Administration” (January 2026); Phillips Lytle LLP, “Mental Health Parity — Past, Present, Future” (April 2026); MoneyGeek, “Mental Health Parity Laws by State 2026”; 2026 Chatbots and Mental Health Survey; NPJ Digital Medicine, chatbot meta-analysis (March 2026); Baylor University Medical Center / PMC, “Too Good to Be True?” (2026); MDPI Encyclopedia, “Digital Mental Health Post COVID-19” (January 2026); UNC Institute for Global Health; BJPsych International (January 2026)
