AI Therapy Laws in 2026: What Your Practice Website Can and Can't Say

States spent 2026 drawing a line between AI as a tool and AI as a therapist. Your website should sit clearly on the right side of it.

SrikaantHOctober 19, 20264 min read

ai laws
ai scribe
informed consent
chatbot
compliance

Open notebook and pen beside a closed laptop on a desk.

Key takeaways

At least seven states now restrict AI therapy, and more bills are coming.
The laws mostly target companies selling chatbots as therapy. But they shape what your own website should say too.
Never describe any tool on your site as a therapist, counselor or "AI therapy."
If you use an AI note-taker, tell clients and get their consent in writing.
A short "How we use AI" note builds trust and keeps you ahead of the next law.

What actually happened this year

A year ago, Illinois was the odd one out. It banned AI from delivering therapy without a licensed professional in charge, and Nevada followed. Then 2026 happened.

Tennessee's law, live since July 1, stops any AI from presenting itself as a qualified mental health professional. Maine made AI therapy without a licensed clinician an unfair trade practice (effective July 29). Vermont now bars AI from delivering mental health services on its own. Colorado's law took effect August 12 and goes further, limiting how licensed psychologists, counselors, social workers and MFTs can use AI in their work.

Rhode Island signed three laws in one go in June. One requires chatbots to spot signs of suicidal thinking and point people to crisis lines. Another bans unlicensed AI therapy outright. The third says any provider using AI to document a session has to tell the patient and check the record for accuracy.

And California? Its bill, SB 903, passed both houses by a wide margin. Governor Newsom vetoed it, saying the definitions were so broad they'd catch legitimate clinical tools. So California has no AI therapy ban for now. Thing is, a near-unanimous vote tells you where this is heading.

The details differ state to state. The direction doesn't: AI can help a licensed person, but it can't be the licensed person.

Does any of this apply to a solo therapist?

Mostly these laws are aimed at apps and platforms. You're not building a therapy bot.

But your website talks about AI more than you might think. A chatbot widget. A line about "AI-assisted notes." A practice-management platform nudging your clients toward an AI coach between sessions. Each one is a place where loose wording can suggest something no law (and no licensing board) wants to see.

This is not legal advice, and rules change fast. Check the specific law in each state where you're licensed, or ask your board.

Words to take off your site

Read your pages with fresh eyes and look for these:

"AI therapist," "AI counselor" or "AI therapy," anywhere
A chatbot with a human name and no label saying it's software
Anything suggesting a tool can assess, diagnose or recommend treatment
"24/7 support" from a bot, which reads like clinical care
Claims that AI "personalizes your treatment"

Swap them for plain descriptions. "Our booking assistant answers questions about fees and availability." That's honest, and nobody can misread it.

If you use an AI note-taker

This is the one most clinicians actually touch. Ambient scribes and AI summaries are everywhere now, often switched on by default inside your EHR.

Rhode Island already requires notice and a clinician's review of the record. California's vetoed bill would have required clear disclosure and affirmative consent before recording or transcribing a session. Expect that idea to come back, there or elsewhere.

So get ahead of it:

Name the tool in your informed consent paperwork and say what it does.
Ask for written consent, and make saying no easy, with no effect on care.
Mention it in your privacy policy, including where recordings go and when they're deleted.
Confirm the vendor signs a business associate agreement.
Review every AI-written note before it's final.

Chatbots: the floor just got higher

We covered the basics in AI Chatbots on a Therapy Website: Where the Line Is. The 2026 laws mostly confirm that advice. A bot should say it's a bot, stick to logistics, show crisis resources and hand off to a human. Rhode Island has now written the crisis part into law.

If your bot can't do all four, switch it off until it can.

Add a short "How we use AI" note

Personally, I think this is the easiest trust win available right now. Clients have read the headlines. Many are wary.

A few honest lines on your FAQ or privacy page do the job: which tools you use, what they never do, that a licensed human makes every clinical decision, and how to opt out. It takes ten minutes to write. And it answers the question a nervous new client won't always ask out loud.