HCPC is silent on AI. CSP has 46 principles.

Open the HCPC's Standards of conduct, performance and ethics, search for "artificial intelligence", and you get no results. Open the Chartered Society of Physiotherapy's AI principles and you get 46 of them, grouped into seven areas. If you run a physiotherapy practice and you are trying to work out what you must do before switching on an AI scribe, a triage tool or an automated outcome-measure scorer, those two facts point in opposite directions, and only one of them can end your registration.

The confusion is understandable. Both bodies appear on the same renewal pages, both are quoted in vendor sales decks, and both are described as "the guidance". They are not the same kind of thing. This explainer sets out which document binds you, what it actually says, what the CSP adds on top, and the one gap that neither body has closed. The HCPC half applies equally to practitioner psychologists, occupational therapists and the other HCPC-regulated professions in your practice.

The document that binds you does not mention AI

The Health and Care Professions Council is the statutory regulator. It regulates 15 health and care professions in the UK, including physiotherapy (HCPC, July 2026). Every registered physiotherapist must meet its Standards of conduct, performance and ethics, and the current version took effect on 1 September 2024 (HCPC, Standards of conduct, performance and ethics, 2024). We checked the full text for this article. It contains no reference to artificial intelligence or emerging technologies.

That absence does not create a loophole. Four standards already do the work. Standard 3.1 says you "must only practise in the areas where you have the appropriate knowledge, skills and experience to meet the needs of a service user safely and effectively." Standard 4.1 says you "must only delegate work to someone who has the knowledge, skills and experience needed to carry it out safely and effectively", and 4.2 says you "must continue to provide appropriate supervision and support." Standard 5.1 says you "must treat information about service users as confidential." Standard 10.1 says you "must keep full, clear and accurate records for everyone you care for, treat or provide other services to."

Applying those four to AI is our reading, not an HCPC statement, but it is the natural one. An AI scribe that drafts your treatment note is producing a record you remain responsible for under 10.1. A tool that suggests an exercise progression is work you have in effect delegated, so 4.1 and 4.2 ask whether you understand it well enough to supervise it. A tool that sends session audio to a third-party server engages 5.1.

So what for you: the rule that can lead to a fitness-to-practise concern is already in force and already covers AI use. Check each tool in your practice against 3.1, 4.1, 5.1 and 10.1 before you worry about any AI-specific document.

The HCPC has promised AI guidance, not published it

On 23 July 2026 the HCPC announced research with Ipsos UK into whether and how registrants use AI across the 15 professions it regulates, starting with a registrant survey in early August (HCPC, July 2026). The same announcement says the HCPC plans to publish guidance on how registrants can "use new technologies such as AI safely and ethically in line with our standards of conduct, performance and ethics." At the time of the announcement, no findings were available.

The only other recent HCPC output we found is a joint statement on AI in health and care professional education, published on 9 February 2026 with the General Osteopathic Council, General Optical Council, Royal College of Veterinary Surgeons and General Pharmaceutical Council (HCPC, February 2026). It covers academic integrity and AI literacy for education providers, not day-to-day practice. When we searched on 3 October 2026 we found no published HCPC practice guidance on AI. Check the HCPC site again before you rely on that, because a regulator can publish at any time.

The wording of the July promise matters. The HCPC says its guidance will apply its existing standards to AI rather than add new duties. That points to a future in which the four standards above stay exactly where they are and a guidance document explains what they mean for a clinician with an AI scribe on the desk.

So what for you: do not wait for that guidance before tightening your own position. When it arrives it will interpret standards you are already held to, and a practice that has already documented its checks will have little to change.

What the CSP's 46 principles add

The CSP is the professional body for physiotherapists, not the regulator. Its AI principles, first announced on 31 March 2025 and last reviewed on 15 November 2025, cover 46 principles across seven areas: general use (9), service delivery (6), clinical practice (7), research and innovation (6), education (6), leadership (4) and the workplace (8) (CSP, AI principles in action). The CSP committed to review them every six months over two years, so expect revisions.

The page does not state whether the principles are binding. Its opening says they were "created to guide members, and those employing physiotherapists on the ethical use of AI in their setting", and as a membership body the CSP has no power to sanction your registration. They apply to NHS and private practice alike.

They are far more specific than the HCPC. On accountability, a physiotherapist "should remain accountable for its application in care", and there "should always be a clear route for human review." On consent, "patients should provide informed consent before AI tools are used in their care", and that consent must be "meaningful, not hidden in fine print." On data, you should "understand how data is stored, processed and shared, and be able to explain this to service users." The practice-level principles also include avoiding vendor lock-in, which matters if your notes sit inside a single scribe vendor's platform.

For a practice owner, the employer-facing principles are the ones most often missed. The CSP asks services to involve staff and trade unions early, and to plan for business continuity if an AI tool fails. Those sit alongside the clinical principles rather than under them.

So what for you: use the CSP principles as your checklist for what a defensible AI policy contains, particularly the consent and data-explanation points, and use the HCPC standards as the reason that policy is not optional.

The gap neither body has closed: your insurance

The line most practice owners have not read sits near the end of the CSP's page: "The CSP PLI policy does not currently specifically address AI" (CSP, AI principles in action, last reviewed November 2025). The CSP notes that liability for AI-related harm could involve app developers, clinicians, patients and manufacturers. HCPC registration renewal also requires you to confirm that you hold appropriate professional indemnity insurance (CSP, HCPC re-registration guidance, 2026).

Put those together and a specific question follows. If an AI-assisted note contains an error that contributes to a patient complaint, will your indemnity policy respond in the way you assume? Neither the HCPC nor the CSP can answer that for you. Your insurer can, in writing.

Frankly, if you have rolled out any AI tool that touches clinical records, the evidence points toward asking your indemnity provider one direct question this month: "Does my policy cover claims arising from an AI-assisted record or recommendation, and are there conditions on the tools I use?" Keep the reply on file next to your tool register. We have covered the equivalent question for psychology in our explainer on professional indemnity and AI scribes.

So what for you: the weakest point in most practices' AI position is not consent wording or tool choice. It is an unanswered insurance question that takes one email to settle.

Where the CQC fits if you are registered

If your practice carries out a CQC-regulated activity, a third body has a view. The CQC published its role and expectations for AI on 21 May 2026, setting out eleven principles that include human oversight, transparency and choice for people using services, staff training, data protection impact assessments and clear accountability for AI-related harm (CQC, May 2026). It also states that it does not assess or approve specific technologies. A regulator that will not approve your tool leaves the evidence of safe use to you. Our CQC AI governance checklist maps those expectations to a one-page practice document.

So what for you: if CQC inspects you, "the vendor said it was safe" will not satisfy anyone. A written register of tools, checks and named owners will.

A four-step position you can document this month

Here is the minimum that satisfies the binding standard, borrows the best of the advisory one, and closes the insurance gap. It is a working method, not regulatory advice, and you should take your own professional or legal guidance for specific cases.

1. List every AI-enabled tool that touches a patient record, appointment or message, including features inside software you already pay for.
2. For each tool, write one line against HCPC standards 3.1, 4.1, 5.1 and 10.1 saying who checks the output and how.
3. Adopt the CSP's consent and data-explanation principles as your patient-facing wording: say plainly that AI is used, what for, where the data goes and what the patient can decline.
4. Send the indemnity question to your provider and file the answer.

That sequence takes an afternoon for a small practice and produces something an HCPC panel, a CQC inspector or an insurer could read and recognise as proportionate.

So what for you: the practices that struggle later are rarely the ones that adopted AI. They are the ones that adopted it with no written record of who checked what.

Frequently asked questions

Is CSP AI guidance legally binding? No statement on the CSP page says so, and the CSP is a professional body rather than the regulator. The binding standards are the HCPC's.

Does the HCPC have AI guidance for physiotherapists? Not yet, as far as we could find on 3 October 2026. It announced research and planned guidance on 23 July 2026, and its current standards contain no mention of AI.

Do I need patient consent to use an AI scribe? The CSP principles say informed consent should be obtained before AI tools are used in care. Separately, UK data protection law and your confidentiality duty under HCPC Standard 5.1 apply to how the data is handled. Take advice on your specific tool.

Does my indemnity cover AI mistakes? The CSP states that its PLI policy does not currently specifically address AI. Ask your provider in writing.

So what for you: if you act on only one answer above, make it the insurance one.

If you want an independent view of which tools in your practice carry the most governance risk, the AI Opportunity & Growth Assessment includes a tool-by-tool review, and you can book a free 20-minute call to talk through where to start.