How States Are Drawing the Line Around AI Mental Health Tools
Artificial intelligence is already being used for questions that once would have been directed primarily to search engines, clinicians, or other health resources. Mental health is part of that shift.
A 2026 KFF poll found that 16 percent of U.S. adults had used AI for mental health information or advice in the previous year. Among adults ages 18 to 29, that figure reached 28 percent. KFF reported.
States are beginning to establish clearer boundaries around AI mental health tools, particularly when technology moves from providing support into functions traditionally associated with clinical care.
The response has not produced one regulatory model. Illinois, Nevada, Colorado, Utah, Rhode Island, and other states have approached the issue differently. Some have restricted AI from independently providing therapy or presenting itself as a mental health professional. Others have emphasized privacy, disclosure, patient consent, and professional oversight. A recent KFF analysis outlines these differing approaches.
Taken together, these laws suggest that the central policy question is becoming more specific. States are not simply deciding whether AI belongs in behavioral health. They are beginning to decide what AI should be allowed to do once it gets there.
What Are AI Mental Health Tools?
AI mental health tools are artificial intelligence systems used to provide mental health information, interact with users, or support behavioral health professionals. Their functions can range from administrative work and documentation to conversational support and treatment-related recommendations. State laws increasingly distinguish between supportive uses and functions considered psychotherapy or other professional care.
That distinction matters because the term covers very different technologies.
An AI system used by a therapist to help schedule appointments is not performing the same function as a chatbot responding directly to someone experiencing psychological distress. A documentation system that organizes clinical notes is also different from a system generating a treatment plan or recommending a therapeutic intervention.
States are increasingly writing those functional differences into law.
Illinois, for example, expressly allows licensed professionals to use AI for administrative and supplementary support when the professional remains responsible for the system's interactions, outputs, and data use. Its definition of administrative support includes functions such as appointment scheduling, billing, insurance claims, and general logistical communications. Illinois law provides the details.
The question is therefore becoming less about the presence of AI and more about the role assigned to it.
How States Are Regulating AI in Mental Health Care
One of the clearest regulatory approaches is to protect activities that states consider professional therapy.
Illinois enacted the Wellness and Oversight for Psychological Resources Act in 2025. The law prohibits AI from making independent therapeutic decisions, directly engaging clients in therapeutic communication, generating treatment recommendations or plans without professional review, or detecting emotions or mental states. Licensed professionals may still use AI within the permitted support functions established by the law. The Illinois statute sets out those limits.
Nevada approaches the issue through the broader category of professional mental or behavioral health care. Its law restricts AI systems specifically programmed to provide services that would amount to professional mental or behavioral health care if a person performed them. It also restricts representations suggesting an AI system is a therapist, counselor, psychiatrist, physician, or similar professional. Administrative support systems used by behavioral health providers remain possible under the law. Nevada's statutory framework provides this broader behavioral health definition.
Colorado has also adopted restrictions surrounding AI and psychotherapy. HB 26-1195 was signed by the governor in June 2026 after passing both legislative chambers. The framework regulates how AI may be used around psychotherapy while preserving a role for permitted professional, administrative, educational, and related uses. Colorado HB 26-1195 outlines that approach.
These laws differ in language and scope. They nevertheless point toward a common distinction.
Regulators appear more comfortable with technology assisting care than with technology independently assuming the authority associated with providing care.
The Emerging Boundary Is Between Support and Clinical Care
That distinction may become one of the more important questions for behavioral health organizations evaluating AI.
Scheduling is relatively easy to classify as administrative. So are billing and many routine workflow tasks.
The boundary becomes less clear when a system begins analyzing a patient's information, communicating directly with that patient, suggesting a course of action, interpreting emotional states, or generating recommendations that could influence treatment.
Illinois provides a useful example. Its statute does not prohibit all AI use by therapists. Instead, it separates permitted support from activities involving independent therapeutic judgment. A clinician can use certain AI functions while remaining responsible for the output, but the system cannot independently make therapeutic decisions or conduct therapeutic communication. Illinois codified those distinctions.
Nevada makes a similar distinction between administrative support and an AI system designed to provide an experience that would constitute professional mental or behavioral health care if delivered by a person. The Nevada statute reflects that division.
Those boundaries are reasonably clear at the extremes. An appointment reminder and independent psychotherapy are not difficult to distinguish.
Many emerging behavioral health applications sit somewhere between them.
Patient check-ins, screening tools, progress monitoring, automated engagement, risk identification, personalized recommendations, and conversational support can combine administrative and clinical elements. As the technology becomes more capable, simply labeling a product an administrative tool may say less about its actual role.
For treatment organizations, function may become more important than category.
Privacy and Consent Are a Separate AI Mental Health Issue
Clinical authority is only part of the regulatory question.
Behavioral health information can be highly sensitive regardless of whether the system receiving it is formally providing therapy. Documentation assistants, transcription systems, chatbots, and other AI tools may process information about symptoms, relationships, substance use, treatment history, medications, trauma, or other personal concerns.
Some states are addressing that issue separately.
Illinois requires written notice and patient consent under specified circumstances when AI is used as supplementary support and a therapeutic session is recorded or transcribed. The patient must be informed that AI will be used and told the specific purpose of the system. Illinois law includes those consent requirements.
Utah has taken a different approach to consumer-facing mental health chatbots. Rather than broadly prohibiting the technology, the state established requirements specifically for suppliers of mental health chatbots. Utah defines these systems around generative AI that engages in conversations similar to confidential communications with a therapist and is represented, or could reasonably be understood, as designed to diagnose, treat, or improve mental health. Utah's guidance explains those requirements.
KFF identifies privacy and disclosure as a separate regulatory direction developing alongside restrictions on independent AI therapy.
That distinction is important.
A system does not need to practice psychotherapy to create meaningful questions about consent, confidentiality, data use, and patient expectations.
Why Consumer Use Makes the Regulatory Question More Urgent
Regulation is developing while consumers are already using these systems.
KFF found that about one in six adults had used AI for mental health information or advice. Use was substantially higher among younger adults, with 28 percent of adults ages 18 to 29 reporting doing so. The KFF tracking poll provides the underlying data.
Access may be part of the picture.
KFF also reported that one in five adults using AI for health information cited difficulty accessing or affording health care as a reason for turning to chatbots. Its reporting found particularly high use of AI for mental health advice among uninsured adults. A related KFF analysis examined those access factors.
Those findings do not establish that AI use causes people to delay or avoid professional treatment. They do show that these systems are being used in circumstances where access, convenience, cost, and immediate availability matter.
That creates a difficult policy balance.
A tool can provide information quickly without necessarily having the judgment, accountability, context, or professional obligations associated with a licensed clinician. At the same time, restricting technology does not resolve the access problems that may lead some people to use it.
The regulatory challenge is therefore broader than determining whether a chatbot gives correct answers. Policymakers are also deciding what users should reasonably be allowed to expect from these systems.
What These Laws Mean for Behavioral Health Providers
For behavioral health organizations, the emerging state frameworks make the phrase "we use AI" increasingly inadequate.
A more useful evaluation begins with what each system actually does.
- Does the technology communicate directly with patients?
- Does it provide therapeutic guidance or behavioral recommendations?
- Does it generate treatment plans or influence clinical decisions?
- Does a licensed professional review its output?
- Does it record or transcribe patient interactions?
- What behavioral health information does it collect?
- What disclosures and consent processes apply?
- Can administrative functions be separated from clinical functions?
- How does the technology describe itself to patients?
These are not only technology procurement questions. They touch clinical governance, compliance, privacy, informed consent, and professional responsibility.
Illinois illustrates why that distinction matters. The state allows certain AI-assisted functions but leaves responsibility with the licensed professional and prohibits specific forms of independent therapeutic activity. The Illinois framework describes those responsibilities.
Nevada similarly preserves administrative uses while restricting systems designed to provide professional mental or behavioral health care themselves. Nevada's statute reflects that distinction.
The practical implication is that organizations may need to evaluate AI at the level of individual functions rather than approving an entire platform under a broad label such as clinical AI or behavioral health technology.
Requirements continue to differ by jurisdiction, and organizations should evaluate applicable laws and professional requirements before deploying these systems. This discussion is informational and is not legal advice.
Substance Use Treatment Is Part of the Same Regulatory Conversation
The debate around mental health AI also reaches substance use treatment.
Behavioral health is broader than conventional psychotherapy, and some state laws reflect that scope.
Nevada's framework uses the category of professional mental or behavioral health care and defines it around services involving the diagnosis, treatment, or prevention of mental illnesses and emotional or behavioral disorders. Its statutory framework encompasses several regulated behavioral health professions. Nevada AB 406 provides the statutory language.
For addiction treatment providers, that makes the larger policy question relevant even when a product is not marketed specifically as a mental health chatbot.
AI-supported documentation, patient communication, screening, engagement, and care recommendations can raise many of the same questions about professional oversight, confidentiality, consent, and clinical responsibility.
The important distinction remains what the technology is being asked to do.
Digital tools can support behavioral health services without automatically becoming the provider of those services. Current state laws are beginning to make that boundary more explicit.
Regulation Is Beginning to Define a Role for AI Rather Than a Single Rule
The state landscape remains fragmented.
KFF's August 2026 analysis identified laws or legislative activity in Illinois, Nevada, Tennessee, Vermont, Rhode Island, Utah, Colorado, California, Pennsylvania, and New Jersey. Some approaches restrict independent therapeutic activity. Others focus more heavily on disclosure and data protections. Several combine elements of both.
There is not yet one settled answer to what an AI mental health system is.
It can operate as consumer software. It can function as administrative infrastructure. It can assist a licensed professional. In some circumstances, it can also begin performing activities that resemble treatment itself.
That classification problem is likely to shape the next phase of regulation.
For behavioral health providers, governance may increasingly require examining individual functions rather than asking whether an organization should use AI in the abstract. A documentation tool, patient chatbot, risk model, clinical recommendation system, and treatment-planning assistant carry different responsibilities even when they are built into the same platform.
The relevant question is becoming more precise.
It is no longer simply whether AI belongs in behavioral health care. It is where technology stops supporting care and begins performing work that states consider care itself.
Frequently Asked Questions
```Are AI mental health tools regulated?
Yes, but regulation varies by state. Some states restrict AI from independently providing psychotherapy or presenting itself as a mental health professional, while others emphasize privacy, disclosure, consent, or professional oversight.
Can AI provide therapy without a licensed professional?
Several states now restrict or prohibit AI from independently performing functions considered psychotherapy or professional mental health care. The specific requirements depend on the state and how the system is used.
Can behavioral health providers still use AI?
In many cases, yes. Several state frameworks preserve administrative, documentation, educational, research, or professionally supervised uses while placing tighter limits on independent clinical functions.
What should behavioral health organizations consider before using AI?
Organizations should examine what the system actually does, whether it communicates with patients, whether clinicians review its output, what information it collects, and what state-specific consent, privacy, disclosure, and professional practice requirements apply.
Conclusion
The different approaches states are taking toward AI in mental health care do not point to a simple divide between governments that support the technology and governments that oppose it.
A more consistent pattern is emerging.
Administrative support, documentation, research, and other professionally supervised functions often remain possible. Independent therapeutic decisions, treatment recommendations, representations of professional status, and direct therapeutic communication are drawing closer scrutiny.
Privacy and consent create another layer of responsibility even when AI is not delivering treatment.
For behavioral health organizations, that makes the question of AI adoption more specific than it was even a few years ago. Evaluating a technology increasingly means understanding exactly what responsibility it is being given, what information it can access, what decisions it can influence, and where professional judgment remains in the process.
AI may continue to take on a larger role in behavioral health. State policy is beginning to clarify that a larger role does not necessarily mean an independent clinical one.
About Arcana Recovery
Arcana Recovery examines how digital tools can support behavioral health and addiction recovery while preserving the role of professional relationships, clinical judgment, and structured care. As technology becomes more integrated into recovery services, understanding where digital support fits within responsible treatment remains an important part of that work.
References
- KFF — Different State Regulatory Approaches Reflect Open Questions About AI Mental Health Tools ```
- KFF — Tracking Poll on Health Information and Trust: Use of AI for Health Information and Advice
- KFF — One in Three Adults Have Used AI Chatbots for Health Advice
- Illinois General Assembly — Public Act 104-0054
- Illinois General Assembly — Codified AI and Mental Health Provisions
- Nevada Legislature — AB 406
- Nevada Legislature — 2025 Statutes
- Colorado General Assembly — HB 26-1195
- Utah Department of Commerce — Mental Health Chatbot Guidance ```
Excerpt
States are beginning to draw clearer boundaries around how artificial intelligence can be used in mental health and behavioral health care. Rather than banning AI outright, many new laws distinguish between administrative support, clinical assistance, and functions that resemble therapy or professional treatment. The result is an emerging regulatory framework centered on professional oversight, privacy, consent, and the question of where technological support becomes clinical care.
