States Are Putting It in Law: A Chatbot Is Not a Therapist

This spring, one state after another took up a question that should never have needed a law: can a software company present a chatbot as a mental health professional? Several states have now limited AI therapy in statute, and Pennsylvania has taken one company to court over it.

The pressure behind these laws is real. A study published June 1 in JAMA Pediatrics by researchers at RAND found that 19.2% of Americans ages 12 to 21 had used an artificial intelligence (AI) chatbot for mental health advice, up from 13.1% a year earlier.¹ Among those young people, 42.8% used a chatbot this way at least once a month, and 63.3% had not told anyone about it.¹

New laws: four states drew a line this spring

Tennessee moved first. Senate Bill 1580, signed April 1 and in effect July 1, bars anyone who builds or deploys an AI system from presenting it as a qualified mental health professional. Violations fall under the state's consumer protection law, with a $5,000 civil penalty for each one and a private right to sue.²

Maine followed on April 13 with LD 2082, which says AI may not deliver psychotherapy. Clinicians in Maine may still use AI tools as a supplement, and only with the client's consent.³

Colorado's House Bill 26-1195, signed June 3, sets rules for the clinical side. AI may take part in therapeutic conversation only in real time with the clinician present, treatment plans need a clinician's approval, recording a session with an AI tool requires written consent, and claiming that an AI provides psychotherapy is now a deceptive trade practice.⁴ Vermont's governor signed H.816, the state's own law on AI in mental health services, on June 17.⁵

Lawsuits: what happens when a bot plays clinician

On May 5, Pennsylvania sued Character Technologies, the company behind Character.AI. The state says chatbots on the platform presented themselves as licensed psychiatrists and that one gave a Pennsylvania license number that is not valid.⁶ Pennsylvania calls that the unlicensed practice of medicine and has asked the court for a preliminary injunction.

A lawsuit filed June 12 against OpenAI alleges that ChatGPT affirmed a 24-year-old woman's suicidal thoughts before her death in July 2025 and never alerted anyone.⁷ It joins a group of product liability cases against OpenAI that a San Francisco court coordinated in February.⁸ None of these claims has been decided, and we expect the companies to fight them. Even so, the pattern they describe is what we would expect from a product designed to keep a person talking with no licensed professional responsible for what it says.

Why it matters: a license comes with accountability

A licensed therapist owes a duty of care to every client, answers to a state board, and can lose the right to practice for causing harm. A chatbot carries none of those obligations. When a company markets one as "support" or a "guide," a person reaching out late at night may not understand the difference, and the RAND findings suggest many young people are not telling anyone who could help them sort it out.

The business incentives point the same way. Large employer and insurer platforms are building AI into their member experience. Lyra Health made its AI guide generally available to members in May, and its announcement included no outcome data specific to the AI.⁹ On May 28, Aetna announced Aetna Mental Health On Demand, a single-session service that uses AI to write clinical notes, starting in January 2027.¹⁰ When software becomes the cheapest unit of care, every pressure inside these companies favors more software and fewer licensed hours.

Therapists feel this too. Each conversation a chatbot absorbs is one a licensed clinician never has, and each benefit design that sends members to an app first teaches people that therapy is something a phone does.

How StartHere.care works

Every therapist a client sees on StartHere.care has at least one license we verified before the profile could go live. Our matching compares what a client tells us with what therapists tell us about their specialties, approach and availability. No chatbot talks to clients, and no AI decides who gets matched.

After the match, we step aside. The client sees the therapists who fit, chooses whom to contact, and works directly with a licensed person. We charge nothing to therapists or clients, so we have no reason to trade clinical hours for cheaper software.

If you are in crisis right now, call or text 988 to reach the 988 Suicide and Crisis Lifeline.

Looking for a licensed therapist who fits how you want to be helped? Start your search here.

Sources

1. Psychiatric News report on McBain et al., JAMA Pediatrics, June 2026.
2. Healthcare Law Insights on Tennessee Senate Bill 1580, April 2026.
3. Maine Legislature, LD 2082, signed April 13, 2026.
4. Colorado General Assembly, HB26-1195, signed June 3, 2026.
5. Vermont Legislature, H.816 (Act 156), signed June 17, 2026.
6. Pennsylvania Governor's Office, lawsuit against Character.AI, May 5, 2026.
7. International Business Times on the June 12 lawsuit against OpenAI, June 2026.
8. Hunton Andrews Kurth on the coordinated OpenAI product liability cases, February 2026.
9. Lyra Health announcement of its AI guide, May 5, 2026.
10. CVS Health announcement of Aetna Mental Health On Demand, May 28, 2026.