Mental Health Practitioners

Mental health practitioners covers private-practice offices of psychologists, social workers, licensed counselors, marriage and family therapists, and other licensed mental-health clinicians who are not physicians. NAICS 62133 specifically covers solo practices, group practices, and employer-sponsored EAPs delivering outpatient psychotherapy and counseling. The market expanded enormously after COVID-19 — telehealth now dominates intake and ongoing therapy. Insurance reimbursement is the central economic constraint: in-network fee schedules for CPT 90834 and 90837 typically run $80-150 per session, while cash-pay practices charge $150-300+. Workforce shortages are acute, with roughly 50,000 unfilled licensure-track positions nationally and worse rural coverage. Group practices (LifeStance Health, Talkiatry, Mindpath Health, Headway, Alma) have rolled up regional fragmentation and now dominate insurance contracting. Combined US spend on counseling and therapy services exceeds $80 billion annually.

The consumer-AI mental health interface drew direct policy and product attention in May 2026. Pennsylvania filed suit against Character.AI in early May over a chatbot posing as a licensed psychiatrist — pushing the entire chatbot industry to harden mental-health-adjacent conversation handling. On May 7, 2026, OpenAI introduced an opt-in Trusted Contact safeguard for ChatGPT — adult users can name a friend or family member to receive a check-in nudge if OpenAI's safety system detects suicidal ideation in a conversation, with under-one-hour human review and no chat content shared in the alert. The feature is opt-in and adult-only by design and does not block users from creating multiple accounts to bypass it; it is a guardrail for users who choose to enroll, not a population-level safety net. Both events signal a shift toward shared responsibility models — chatbot vendors building self-harm detection into the consumer surface itself, while licensed mental-health practitioners remain the clinical layer for diagnosis, treatment, and crisis intervention.

Wysa is the most-deployed mental-health AI chatbot — clinically validated CBT-based assistant used by employers, health plans, and therapy services for first-line triage, sleep hygiene, and self-guided exercises. Spring Health, Talkspace, and Headway use AI for clinician matching, scheduling optimization, and outcome tracking across their large therapist networks. AI scribes are creeping into therapy documentation despite session-confidentiality concerns — Abridge, Suki AI, and Dragon Copilot all offer mental-health-tuned variants. Predictive risk models flag suicide risk, dropout risk, and treatment non-response in real time, with internal models running at LifeStance, Spring Health, and major academic medical centers. Population-health platforms (Innovaccer, Palantir AIP) help health plans identify members who would benefit from earlier mental-health intervention. The horizontal foundation models (ChatGPT, Claude, Microsoft Copilot) have become daily research aids — clinicians use them to draft treatment plans, look up DSM criteria, and prepare for licensure exams.

In May 2026, Pennsylvania filed suit against Character.AI after state investigators interacted with a chatbot named "Emilie" that claimed to be a licensed psychiatrist and produced a fabricated state medical license number while discussing depression treatment. The complaint cites Pennsylvania's Medical Practice Act and is the first state action specifically targeting AI chatbots that impersonate licensed mental-health professionals — distinct from Kentucky's January 2026 Character.AI suit, which focused on harm to minors rather than credentialing fraud. The Pennsylvania filing creates a credentialing-fraud precedent that has direct implications for how mental-health AI products represent themselves to users; expect more state attorneys general to follow the same template, and expect platform-level disclosures and identity-verification flows to harden across consumer mental-health AI in response.

The biggest workforce question is whether AI chatbots are substitutes or complements to human therapists. Evidence so far points to complements — chatbots handle subclinical anxiety, sleep concerns, and CBT-skills practice that would otherwise go untreated, while humans handle clinical-grade work. But the line blurs at the edges. Insurers are quietly funding AI-first triage as a way to ration access in an undersupplied market, which is promising for access but troubling for clinical quality if oversight is light. Documentation automation is unambiguously freeing therapists from 10-15 hours per week of administrative work — often the difference between burnout and sustainability. The biggest open risks are regulatory: the APA, state licensure boards, and the FDA are all developing AI guidance, with significant uncertainty about which AI tools require medical-device clearance versus wellness-product framing. Therapists who learn to use AI as a documentation and research aid will gain meaningful capacity; those who do not will lose competitive ground.