Counsellor / psychotherapist

A Colorado statute binding everyone licensed or permitted to provide psychotherapy in the state, signed on 3 June 2026 and effective from 12 August 2026. It defines 'supplementary support' that AI may provide, including preparing and maintaining client records, including therapy notes. It provides that a practitioner shall not use an AI system to record or transcribe a client's therapeutic session unless the client is informed in advance in writing that AI will be used and for what purpose, and consents in writing; a refusal or later revocation shall not be used as a basis to deny psychotherapy services. A separate subsection exempts tools that provide things such as therapeutic session summaries if they do not diagnose or treat, which may take some tools outside the consent rule. One state; enforcement runs through the licensing boards, and the text does not say how many therapists use such tools.

Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.

A Maine statute covering counselling professionals, with the same wording repeated for psychologists, social workers and others, approved by the Governor on 13 April 2026. Before AI is used in supplementary support, the client must be informed in writing that AI will be used, of its specific purpose, and of how session data collected by AI will be stored, retained, used for training and deleted when therapy ends, and must consent. A licensee may not allow AI to make independent therapeutic decisions, to interact directly with clients in any form of therapeutic communication — defined to include offering emotional support, reassurance or empathy — or to generate treatment plans without the licensee's review and approval. The licensing boards still have to adopt implementing rules; it is one state.

Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.

England only, and only products entering NHS Talking Therapies for anxiety and depression — private apps and other health systems are outside it. What it establishes is a gate, not a deployment: the page says the assessment "gives services more confidence in the technologies they consider, select, commission and use locally", so local services decide. Read the conditions rather than the counts: Koa Health Perspectives was recorded as not compliant because it "is designed to deliver a low-intensity equivalent treatment" for a condition its low-intensity practitioners are not trained in, and three 2025 products are compliant only "provided the product is routinely delivered with the support of a high-intensity CBT therapist". Of the chatbot product assessed, Wysa for generalised anxiety, NHS England wrote that it is "compliant with the core and clinical content criteria" but that "there is not yet sufficient research demonstrating impact". The page carries no publication date of its own, and NHS England does not publish the individual assessment dates; the date here is the earliest Internet Archive capture holding the 2025 batch (2025-10-08), and the previous capture, 2025-06-15, does not, so the outcomes became public between those two dates. The page also still says "Six products have now been assessed" while listing eleven.

Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.

One US state, announced by the state agency that regulates the professions concerned. It is regulation arriving ahead of deployment rather than after it, which is unusual on this site. It does not bind other markets, and it does not reach products marketed as wellness or coaching rather than therapy — which is where most consumer products currently sit. A law also states what may not be offered, not what people actually use.

Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.

Read on the Digital Agency's own statute portal, in the version in force from 1 June 2025; the Act dates from 2015. Two provisions bear on this page and they work differently. Article 42(2) is a duty on the practitioner: where a person needing psychological support has an attending physician for that support, the certified psychologist shall take that physician's instruction. Article 42(1) separately requires the practitioner to maintain coordination with those providing health, welfare and education services. Article 44 protects the title — a person who is not certified may not call themselves 公認心理師, nor use the characters 心理師 within any name they use — and article 49 makes breaching that punishable by a fine of up to ¥300,000, so the title restriction is criminal rather than professional. What this does not do is regulate any technology: the Act predates the services this page is about and names none of them. It constrains who may hold themselves out under this title and whose instruction binds a therapeutic decision, which is a fact about the professional structure rather than about what software can do. It applies in Japan and counts nobody.

Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.

A lab study that mapped therapy guides used by major medical institutions — for example the importance of a therapeutic alliance — and tested current language models such as gpt-4o against them. The models expressed stigma toward those with mental health conditions and responded inappropriately to certain common and critical conditions, including encouraging clients' delusional thinking, even in larger and newer models; the authors also point to barriers such as a therapeutic alliance requiring human characteristics. It tests models in constructed scenarios, not products or counsellors.

A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.

A randomised trial of 106 US adults with major depressive disorder, generalized anxiety disorder or an eating disorder, who used a smartphone chatbot by responding to prompts or initiating conversations when they needed to talk. Depression symptoms fell 51% on average and anxiety 31%, and participants reported trusting the system to a degree comparable to working with a mental health professional; the investigators say no generative AI agent is ready to operate fully autonomously in mental health. The comparison was with people without the chatbot, not with counsellors, and the lab built the software.

A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.

The part of the market that is easiest to enter — brief, low-intensity, high-volume support — is the part where a free product is available at three in the morning and does not make anyone feel judged. That is worth knowing before you plan around it. The parts that hold are the ones you cannot practise without supervision and hours: risk, depth, and being a specific person over months.

Regulation arrived here ahead of deployment, which is rare on this site and worth using: at least one jurisdiction has placed therapeutic decision-making with a licensed person by statute, and named administrative support as the permitted use. The practical question is not whether products will exist — they do — but whether the ones your clients use call themselves therapy or wellness, because only the first is covered by that line.

Stay and strengthen

Work where risk is part of the job

Settings that handle risk — crisis, safeguarding, complex presentations — are the ones where a legal line already places the decision with a licensed person, and where the low-intensity substitute does not reach.

Reshape the role

Be the one who decides which tools may be in the room

Where regulation permits administrative and supplementary uses, somebody has to judge which side of that line a given product sits on, and what clients are told. In most services nobody has been given that.

Adjacent move

Move to supervision and training

If entry-level, low-intensity work thins, the people entering the profession still have to be trained and supervised by someone — and that requirement is written into licensure rather than into demand.

The parts move in opposite directions here more sharply than on most pages, so one number would be actively misleading. The lightest, highest-volume end of the market now has a free substitute that is available at three in the morning. Therapeutic decision-making has gone the other way: at least one jurisdiction has placed it with a licensed person by statute. A signal you can check yourself: of the people who contacted your service last month, how many were in the lightest tier — and how many of them had already tried something else first.

This site cannot answer that clinically and does not try to — it is about work, not treatment. What it can tell you is what a law says: in August 2025 Illinois enacted a statute prohibiting AI from providing mental health and therapeutic decision-making, permitting it only for administrative and supplementary support. That is one US state, it does not bind anywhere else, and it does not cover products marketed as wellness or coaching rather than therapy — which is where most of them sit. If this question is about your own care, it is one for a practitioner rather than for a website.

Worth separating the two things being compared, because they are not the same product. What is always available is a conversation that responds to what you bring it. What a practitioner is paid for is noticing what you did not bring — the pattern across months, the subject you keep moving away from — and deciding, and being answerable, when something turns out to be more serious than it sounded. The first is genuinely valuable and genuinely cheap now. The second has not moved, and in at least one jurisdiction it is now legally required to be a person.

This site does not tell anyone which training to do, and a page that did would be worth less. What it can give you is the shape: the entry tier of this market is where a free substitute is strongest, and the tiers that hold — risk, depth, supervision — are the ones that require hours you can only accumulate by practising. That is an uncomfortable order, and it is the same order this site finds in law, audit and medicine. Ask the training body a specific question: where do their graduates get their first supervised hours, and has that changed.