AI for Therapists — the CE Courses Worth Paying For, and the Ones That Aren't

AI for therapists in 2026 means two courses, not twenty. Person Centered Tech’s Beyond Hype and Anxiety ($119, 4 legal-ethical CE hours) is the one to take: it covers PHI, de-identification, informed consent, and state law. Add its $49 session on AI documentation for the note-quality analysis. Skip generic AI certificates: no licensing board counts them. Then pick one AI scribe with a signed BAA. Five states now restrict AI from delivering therapy; Colorado’s rules start 12 August 2026.

The course list is short because only one thing about AI can cost you your license, and it isn’t prompting skill.

What AI for therapists actually changes in the week

Four places, and only four. Everything else is noise.

The session note

This is the whole ballgame. An AI scribe records or transcribes the session, then drafts a SOAP or DAP note in about a minute. That’s why ai therapy notes is now a $56-CPC search term, and why ai for therapists notes costs $43 a click. The vendors are bidding hard for you.

The AI tools for therapists that actually matter are these three, priced off their own pages on 6 August 2026. Upheal charges $1 per session, capped at $69/month, after a 30-day trial. Mentalyc runs $19.99/month for 40 notes up to $119.99 for 330 on monthly billing; commit to a year and the same tiers drop to $14.99 and $99.99. Blueprint bills per session at three levels, and its entry tier is the cheapest AI documentation named on this page: $0.49 for a scribe that sits beside the EHR you already have, $0.99 to add clinical decision support, $1.49 for Blueprint’s own EHR with AI chat. All three publish HIPAA and SOC 2 compliance, and all three sign a BAA.

If you want the single answer: the best AI for therapists doing straight documentation is Blueprint at $0.49 a session, because per-session billing means a light caseload never subsidises a heavy one. Upheal wins if you want session insights rather than just a note; its cap bites at 69 sessions, but against Blueprint’s $0.49 it only works out cheaper past about 140 sessions a month, which is a full-time caseload and then some. Mentalyc is the pick if you bill insurance and want the note shaped for a payer from the start. Any of the three beats writing notes at 9pm.

Each one has a free door, and each door is narrower than the marketing suggests. Upheal’s free plan does include unlimited notes, but only in SOAP, DAP and intake formats, and a call held on Upheal’s own telehealth won’t produce a note unless you’re paying. Mentalyc gives you 14 days without a card. Blueprint’s free Core tier is the EHR with the scribe switched off; AI notes start at $0.49 a session.

Two rules the tools won’t enforce for you. Sign the BAA before the first recorded session. And read every draft for medical necessity language, because a scribe writes what it heard, not what justifies the claim. That gap is what auditors find.

Treatment plans and assessments

Upheal and Mentalyc both generate treatment plans; Blueprint threads validated measures like the PHQ-9 and GAD-7 into the documentation flow so outcome data lands in the note instead of a drawer. Legally this is the delicate one: an AI-generated treatment plan entering the chart without your substantive review is exactly what Illinois wrote a statute about.

The admin layer nobody certifies you for

Scheduling, reminders, superbills, claims, the intake form nobody filled in. This is where the hours actually are, and no CE course teaches it. To build these flows rather than rent them, our free AI automation course is the practical route.

The client who brings ChatGPT to session

New this year, and the one your training probably didn’t cover. Clients arrive having already processed the week with a chatbot, attached to its interpretation, transcript on their phone. Person Centered Tech built a whole module on it. Treating it as an intrusion is a mistake; it’s clinical material.

The 3 AI courses for therapists worth taking in 2026

We compared the AI CE offerings carrying real board approvals, verified their prices on the provider’s own page (PESI and all three scribes re-checked 6 August 2026; the two Person Centered Tech prices last read 30 July), and ranked them by one test: does this change what you do on Monday, and does it count toward renewal?

1. Person Centered Tech — Beyond Hype and Anxiety

$119 · 4 legal-ethical CE hours · on-demand self-study · APA + NBCC (ACEP #6582) approved

Nine modules from Dr. Maelisa McCaffrey and PCT director Liath Dalton, and the only course here that treats compliance as solvable rather than scary. It covers what counts as PHI, why informal anonymization is not HIPAA de-identification (PCT’s position: there is no such thing as a de-identified transcript), consent language, practice policy, documentation integrity, and the state-law overlay.

The handouts are the real value: platform-specific prompting guides for SimplePractice, TherapyNotes, Jane, Quill, Berries, Blueprint and Mentalyc, plus an AI vetting questionnaire and a decision tree.

Honest negative: four hours of legal-ethical CE is a lot of regulation and not much clinical technique. You’ll finish knowing how to choose a tool safely, not how to get more out of it.

2. Person Centered Tech — Using AI as a Mental Health Clinician

$49 · 2 legal-ethical CE hours · on-demand self-study · Dr. Maelisa McCaffrey

The cheapest credible option, and the one with the most useful artifact in this space: McCaffrey runs progress notes from multiple AI platforms side by side and grades them against ethical documentation standards and medical necessity. If you’ve wondered whether your scribe’s output would survive an audit, this is the two hours that answers it.

What it isn’t: current. The course predates the 2026 state statutes, and its citations stop in 2023. This is a documentation course with ethics attached, not a legal update, which is exactly why the $119 course exists.

3. PESI — The Ethics of AI in Therapy

$169.99 · 3 hrs 14 min · digital seminar · Joni M. Gilbertson, MA, NCC, LMHC, BCTMH, CTMH

Broader scope than the PCT pair: chatbots, avatar counseling, robots, VR, state and federal law, and three case studies (substance use, depression, OCD). Worth it for the therapy-delivery side rather than just documentation, or if your board already accepts PESI.

The negative here is a big one. It’s the most expensive option on the page and the least clear about what you’re buying: PESI advertises ethics credit but publishes no per-profession CE hour breakdown, and a second participant’s CE test costs another $24.99. The recording is also dated December 2025, so its “latest laws” section cannot contain the three 2026 state statutes below. Call them before you pay.

Clearly Clinical carries a one-hour law-and-ethics episode on AI documentation (Ep. 216, McCaffrey again). The audio itself is free on YouTube and Apple Podcasts; what the membership buys is the credit, which needs an 80% score on the post-test. It carries no ethics credit for New York psychologists.

Compared at a glance

Prices verified on each provider’s own page. PESI, Upheal, Mentalyc, Blueprint and the Elements of AI hour estimate were re-read on 6 August 2026; the Google AI Essentials course length was read on Coursera on 5 August. The two Person Centered Tech prices date from 30 July, because their checkout renders the figure in a script we can’t read. Coursera’s $49/month never renders as text at all, which is why it carries a flag.

Do not bother with a generic AI certificate

If you are a licensed clinician, a Coursera or IBM or MIT AI certificate does nothing for you. Zero boards accept it toward renewal, no client will ever ask, and none of that curriculum mentions PHI, a BAA, or medical necessity. The AI certifications guide exists for people whose employers screen on those credentials. Yours doesn’t.

Skip anything selling a path to building or supervising an “AI therapist,” too. In Illinois that’s a $10,000-per-violation problem.

What the new state laws changed

Illinois’ Wellness and Oversight for Psychological Resources Act (the WOPR Act, HB 1806, Public Act 104-0054, effective 1 August 2025) drew the first clear line in the country. AI may do “administrative support” (the statute’s examples are scheduling and billing) and “supplementary support” such as maintaining records or analyzing anonymized data. It may not make independent therapeutic decisions, engage in therapeutic communication with a client, generate treatment recommendations or plans you haven’t reviewed and approved, or detect emotions and mental states. Before AI touches a session that gets recorded or transcribed you need informed consent that names the specific use and purpose. A blanket clause in your intake paperwork will not do it. IDFPR enforces at up to $10,000 per violation.

Nevada’s AB 406, signed June 2025 and effective 1 July, prohibits offering AI systems that provide professional mental or behavioral healthcare. Read past the headline, though, because Nevada aims at you as well as at the vendors: a Nevada provider may not use an AI system in connection with care delivered directly to a patient at all. Administrative support only.

Then 2026 brought three more. Maine’s LD 2082, signed 13 April 2026 and in force since 29 July, bars anyone from providing, advertising or offering therapy or psychotherapy, AI included, unless a licensed professional delivers it. It also sets the consent bar high: where AI supplies supplementary support on a recorded or transcribed session, the client must be told in writing what the AI is for and how the session data is handled, and must give written, revocable consent. A broad terms-of-service acceptance doesn’t count, and you can’t refuse to see a client who declines. Rhode Island’s Oversight of Artificial Intelligence Technology in Mental Health Care Act (S 2197, effective 22 June 2026) copies the Illinois prohibitions and lands on that same written standard.

Utah took the other road entirely. HB 452, effective May 2025, lets chatbots operate but demands disclosure that the user is talking to a machine before first use, again after any seven-day gap, and any time the user asks, enforced by the Division of Consumer Protection at $2,500 a violation.

Colorado is the one to diary. Governor Polis signed HB26-1195, Psychotherapy Artificial Intelligence Restrictions, on 3 June 2026; it sits in the session laws as Chapter 358 and takes effect 12 August 2026 — days after this page was last verified. Read the enacted text rather than the headlines, because it is narrower and stranger than “Colorado bans AI therapy.” AI may not carry on therapeutic communication with a client unless you, the AI and the client are in synchronous, real-time interaction. It may not generate therapeutic recommendations or treatment plans you haven’t reviewed and approved. Administrative and supplementary support stay legal, with you on the hook for reviewing every output. Like Maine and Rhode Island, it requires written informed consent before AI records or transcribes a session, disclosed in advance along with the purpose, and it adds a duty no other state has: at initial client contact you must inform the client of the prohibitions on AI in psychotherapy. It also makes it a deceptive trade practice to represent that an AI system provides psychotherapy, that its output is equivalent to a regulated professional’s, or that what a user tells it is confidential the way a therapist is.

Note the enforcement mechanism, because it bites harder than a fine. Colorado routes licensee violations to your licensing board at DORA. The exposure is your license, not your bank balance.

Five states, five different drafting instincts, and the list is still moving. Louisiana’s HB 475, enacted 28 May 2026, took the softest route available: tell the patient before an AI transcribes the visit, by signage or intake form or a sentence out loud. No consent, no right to refuse. Assume your state legislates next, and assume it doesn’t copy Louisiana.

The pattern worth reading off all five: written consent is becoming the default, not the strict case. Three of the five already demand it, and they are the three most recent — Maine, Rhode Island, Colorado. Draft your consent language to that standard now and you won’t be rewriting it the month your own state moves.

None of this restricts the scribe on your desk. All of it restricts what you may let a model do unsupervised, and it means “my documentation tool wrote that” is not a defense anywhere.

The genuinely free path

You can get competent on this for nothing.

Start with Person Centered Tech’s free CE course: HIPAA Security Compliance in Mental Health, a guided reading worth 1 legal-ethical CE hour at $0.00, which also comes with their Solo Practice Care forms pack. HIPAA fluency is the prerequisite for every AI decision you’re about to make.

Then take Elements of AI from the University of Helsinki and MinnaLearn. Introduction to AI costs nothing, issues a free certificate, and needs no math or programming; its makers reckon 4–8 hours per part and suggest giving yourself six weeks. The Building AI sequel is free to take and charges only for the certificate. Google AI Essentials is the paid step up: five short courses, under 10 hours all in by Google’s own count, financial aid available, $49/month (verify) once Coursera’s free trial ends. We reviewed it in full.

Finally, test your chosen scribe on real sessions before you commit. Upheal gives you 30 days and then a permanent free plan; Mentalyc gives you 14 days without a card; Blueprint hands you a batch of free sessions to start. That is enough runway to find out whether the notes it writes are ones you’d sign. Our free AI courses hub tracks every no-cost option we’ve verified.

To go past safe tool use — building your own documentation and intake workflows rather than renting someone’s — a structured program is worth the money. Towards AI runs hands-on AI training our readers get a graduate discount on. Take the free CE first.

This page sits under AI courses by profession. Building an AI-literate practice and hiring for it? Roles go up on our talent board.

FAQ

What is the best AI course for therapists in 2026?

Person Centered Tech’s Beyond Hype and Anxiety at $119 for 4 legal-ethical CE hours. It’s APA- and NBCC-approved, covers PHI, consent and state law across nine on-demand modules, and ships platform-specific prompting guides, an AI vetting questionnaire and a state-law overview you’ll actually use afterward. That handout is why it beats the cheaper option.

Are AI courses for therapists eligible for CE credit?

The ones from board-approved providers are. Person Centered Tech (APA sponsor, NBCC ACEP #6582), PESI and Clearly Clinical all issue CE. Generic AI courses from Coursera, Google or MIT issue certificates but no CE credit, and no licensing board accepts them toward renewal.

Is it legal to use AI to write therapy notes?

Generally yes, with conditions. You need a signed BAA with the vendor, client informed consent, and your own substantive review of every note. Illinois’ WOPR Act bans independent therapeutic decisions and unreviewed treatment plans, not documentation support, but requires specific consent before AI touches a recorded or transcribed session. Maine, Rhode Island and Colorado want that consent written.

What AI tools do therapists actually use?

Mostly documentation scribes: Upheal ($1/session, capped at $69/month), Mentalyc ($19.99–$119.99/month by note volume) and Blueprint ($0.49–$1.49/session, depending on how much beyond notes you want). EHRs like SimplePractice, TherapyNotes and Jane now have AI note features built in. All should be run under a BAA.

Can AI replace a therapist?

Five states have answered that legislatively, and the answer is no. Illinois, Nevada, Maine, Rhode Island and Colorado (HB26-1195, effective 12 August 2026) all bar AI from delivering therapy to the public; Utah allows chatbots but forces them to admit what they are. The defensible use of AI for counselors and therapists is documentation and admin, not the therapeutic relationship.