Industry
Medicaid Work Requirements 2027: PMHNP Exemption Guide
Medicaid work requirements start January 1, 2027, and a diagnosis alone won't exempt patients. What PMHNPs must document for the medically frail exemption.
Read More →
TL;DR
- Seven states now restrict AI in mental health care: Illinois and Nevada (2025), plus Colorado, Maine, Rhode Island, Tennessee and Vermont (2026). Utah and California regulate chatbots through disclosure and crisis rules instead.
- Illinois’s law names the licensed advanced practice psychiatric nurse and leaves physicians out. A PMHNP there is covered, and the psychiatrist down the hall is not.
- Scheduling, billing, reminders and drafting notes are still allowed. The laws call this administrative and supplementary support.
- Recording or transcribing a session for an AI scribe takes the patient’s written, specific, revocable consent in Illinois. Patient consent can’t make the banned uses legal.
- Three uses are off limits: AI making therapeutic decisions on its own, AI holding therapeutic conversations with patients, and AI detecting emotions or mental states.
- Illinois fines run up to $10,000 per violation, enforced by the state’s professional licensing department.
In this article
In Illinois, the state’s AI therapy law covers the psychiatric nurse practitioner and leaves the psychiatrist alone. Its definition of a licensed professional lists the “licensed advanced practice psychiatric nurse” and then ends with three words: “except for a physician.”
That detail matters because the Wellness and Oversight for Psychological Resources Act carries civil penalties of up to $10,000 per violation. It took effect in August 2025. Since then, the list of states restricting AI in mental health care has grown to seven.
Most coverage has focused on chatbot companies. The text of these laws reaches further than that. They govern what you, the licensed clinician, may let software do inside your practice: the scribe running during a session, the patient-messaging bot, and the tool that drafts a treatment plan.
The good news is that most of what PMHNPs actually use AI for is still allowed. The risk sits in a few specific features, and some popular tools ship with those features already turned on.
Seven States, Three Models
State AI mental health laws fall into three groups. Which group your state is in decides how much of this applies to you.
Model 1: Ban therapy by AI, carve out the clinician
Nevada’s Assembly Bill 406 took effect in July 2025, and Illinois’s law followed in August. In 2026, Colorado, Maine, Rhode Island, Tennessee and Vermont passed their own restrictions, which brings the total to seven states. Maine’s LD 2082 lists licensed advanced practice registered nurses by name among the professionals it covers.
These laws share a template. AI can’t provide therapy or present itself as a therapist. Licensed professionals can still use it for paperwork, as long as a human owns the clinical care.
Model 2: Disclosure (Utah)
Utah’s HB 452 lets mental health chatbots keep running. In exchange, they have to tell users they are AI and not a human, limit their advertising claims, and protect user data.
Model 3: Crisis protocols (California)
California’s SB 243 took effect January 1, 2026, and covers companion chatbots. It requires them to disclose they aren’t human, detect suicidal ideation and refer users to crisis resources, and add protections for minors. Users can sue over harm.
If you practice across state lines, as more PMHNPs will once the APRN Compact goes live, you need to know all three models. A single AI workflow has to clear the strictest state on your license list.
The Green List: What the Laws Still Allow
Illinois sorts permitted uses into two buckets, and Maine copies the structure almost word for word. Knowing the exact categories keeps you from giving up tools you’re allowed to keep.
Administrative support
This covers tasks that don’t involve therapeutic communication. The statute’s examples are appointment scheduling and reminders, processing billing and insurance claims, and drafting general communications about therapy logistics.
- Automated appointment reminders and rescheduling
- Prior authorization paperwork and claim scrubbing
- Intake packet assembly and form routing
- Welcome letters and practice-policy emails
Supplementary support
This bucket covers maintaining client records, including preparing therapy notes, and analyzing anonymized data to track progress. A licensed professional has to review the output. A 2025 analysis in JMIR Mental Health points out that this category includes the AI note-drafting tools many clinicians already use.
For most PMHNPs, this means the time-savers covered in AI Tools for PMHNPs in 2026 are still allowed. A 20-patient day with ambient documentation can still save an hour or more of charting. What changes is the paperwork you need before you turn the tool on.
Use AI for the paperwork, never for the person. If a tool says something to your patient about their symptoms while you’re not in the room, you own every word of it.
— Lindsay Hill, DNP, PMHNP-BC
AI Scribes and the Consent Line
Most PMHNPs will run into the consent rule before anything else. In Illinois, if supplementary support uses recordings or transcripts of a session, the patient has to give consent first.
The statute defines consent narrowly. It has to be a clear, affirmative act that shows a written agreement, and electronic signatures count. It also has to be specific and revocable. Patients agreeing to general terms of service, hovering over a box, or staying silent don’t count.
A consent workflow that holds up
- Separate form. Don’t bury AI recording consent inside the general intake packet. Use its own page with its own signature.
- Name the function. Say plainly that the session audio will be processed by software to draft a note, and that a clinician reviews every note.
- Revocation path. Tell patients they can say “no recording today” at any visit, and build a one-click way to switch the scribe off.
- Chart the status. Keep a visible flag in the record so a covering clinician knows whether consent is on file.
- Re-consent on changes. If you switch vendors or the tool adds features, get a new signature.
Documentation discipline protects you in more places than the AI law. The same records you would hand a payer during an insurance audit also show a licensing board that a human reviewed and signed each note.
The Red Lines: Three Uses No Consent Can Fix
Illinois bars licensed professionals from letting AI do three things, and a patient’s signature doesn’t change that. Maine’s bill uses nearly the same language.
- Making independent therapeutic decisions. Software can’t choose a treatment direction without you.
- Engaging in therapeutic communication. A bot can’t hold supportive, reflective or symptom-focused conversations with your patients.
- Detecting emotions or mental states. Illinois prohibits AI emotion detection outright.
Maine adds a fourth: AI may not generate therapeutic recommendations or treatment plans unless the licensed professional reviews and approves them. The draft is allowed. Sending it out unreviewed is the violation.
Where real products cross the line
The violation is usually a feature, not the tool as a whole. Watch for between-visit “check-in” bots that answer patient mood reports, mood and sentiment scores generated from session audio, and patient-portal assistants that reply to symptom questions after hours.
Crisis handling raises the stakes. A bot that answers a patient’s 2 a.m. message about wanting to die is doing therapeutic communication at the moment it matters most. Your suicide risk assessment process has to route those messages to a human or a crisis line, not to a model.
What Most People Get Wrong
The common assumption is that these laws target app companies, and a clinician using AI “under supervision” is safe. The statutes say otherwise.
They are written into professional licensing law. That puts enforcement with the boards and departments that oversee your license, not only consumer-protection regulators. Illinois’s penalty is assessed per violation, and an auto-reply feature running across a patient panel can produce a lot of violations.
The second mistake runs the other way: deciding that AI is banned and dropping it entirely. That gives up real time savings for no legal benefit. Scheduling, billing and reviewed note drafts are explicitly protected.
The third mistake is assuming federal action has settled the question. A December 2025 executive order told federal agencies to challenge state AI laws, but state statutes stay in force until a court or Congress says otherwise. The American Psychological Association tells clinicians to track the requirements in every state where they practice, and that advice applies to PMHNPs too.
The new grads I worry about aren’t the ones avoiding AI. They’re the ones who turned on every feature in the free trial and never read what the patient-facing parts do.
— Lindsay Hill, DNP, PMHNP-BC
A 30-Minute AI Audit for Your Practice
Block one half-hour this month and walk through every AI-enabled tool you use, including the features built into your EHR.
- Inventory (5 minutes). List every tool: scribe, EHR assistant, patient portal, texting platform, intake software. Note which ones touch session audio or patient messages.
- Sort (10 minutes). Label each feature administrative, supplementary or red-line. Anything that talks to patients about symptoms goes in the red-line column.
- Switch off (5 minutes). Disable red-line features in states with restriction laws. Turn off sentiment and emotion analytics wherever you can.
- Consent (5 minutes). Confirm you have a standalone, revocable AI recording consent form, and that every active patient using the scribe has signed it.
- Review trail (5 minutes). Make sure every AI-drafted note and treatment plan shows a clinician review and signature before it’s final.
Telehealth-heavy practices should repeat the audit for each state they’re licensed in. The setup advice in our guide to launching a telehealth-only PMHNP practice assumes a clean tech stack, and this audit keeps it that way.
Not sure whether a specific feature crosses the line? That’s the kind of question to bring to a mentor before a board brings it to you. The Psych NP Fellowship gives early-career PMHNPs 12 months of live case and practice-operations support for questions like this.
Frequently Asked Questions
Can PMHNPs use AI scribes under the new state laws?
Yes. In Illinois, drafting notes counts as supplementary support, which licensed professionals may use. If the tool records or transcribes the session, the patient must give written, specific and revocable consent first, and you have to review what the tool produces.
Do AI therapy laws apply to medication management visits?
Probably, if any psychotherapy happens in the visit. Illinois’s law covers therapy and psychotherapy services, and plenty of med-management visits include supportive therapy billed with an add-on code. The safer move is to apply the same consent and review rules to every psychiatric visit.
Can I use an AI chatbot to check in with patients between visits?
Only for logistics. Appointment reminders and general practice messages count as administrative support. A bot that asks how patients are feeling and replies to what they say is therapeutic communication, and it’s banned in the restriction states even if the patient agrees to it.
Which state’s AI law applies to a telehealth visit?
Plan for the patient’s state. Professional practice rules generally follow where the patient is located during the visit, so a PMHNP licensed in several states should build one workflow that meets the strictest law they practice under.
Did the December 2025 executive order cancel state AI laws?
No. The executive order told federal agencies to push back on state AI laws, but an executive order can’t repeal a state statute on its own. Until a court strikes a law down or Congress preempts it, the state law still binds your license.
The Bottom Line
AI therapy laws don’t take AI away from PMHNPs. They draw a line between software that helps the clinician and software that acts like one. Scheduling, billing and reviewed note drafts are allowed. Bots that hold conversations with patients, make decisions on their own or read emotions are not.
The states that have acted so far name psychiatric APRNs directly, and more legislatures will take this up in 2027. The cheapest protection is the one you can do this week: a standalone recording consent form and a 30-minute audit of what your tools say to patients when you’re not there.
Your next step: before your next session, open your scribe’s settings and find out whether it sends anything to patients. If it does, turn that feature off until you’ve checked your state’s law.
Explore the Psych NP Fellowship
A 12-month clinical mentorship for new and early-career PMHNPs. Live case support, practice-operations guidance and a community that has already worked through these questions.
This content is for educational purposes and does not replace individualized clinical judgment or supervision.
About the author: Lindsay Hill, DNP, PMHNP-BC is the founder of the Psych NP Fellowship, a 12-month clinical mentorship program for new and early-career psychiatric nurse practitioners. She is a published contributor to Psychiatric Times, past President of the Arizona APNA Chapter, and co-founder of the Psych NP Network.
