12 August 20267 min read

Can you use AI to write your reflective account? What the HCPC and NMC actually say

It is one of the most common questions clinicians are asking in 2026, and one of the worst served by search results. Most of what comes up is written by companies selling AI tools, which is not the best place to find out what your regulator requires.

So here is the honest position, taken from what the HCPC and the NMC have actually published, with the gaps flagged as gaps.

The short answer

You can use a tool to help you think, to structure what you have already said, and to tidy your writing. You cannot hand over authorship.

The published standards are not about which software you opened. They are about whose work the document is, and whether it is true.

What the HCPC standards actually say

The HCPC has five standards of continuing professional development. Two of them do the heavy lifting here.

Standard 1 requires registrants to "maintain a continuous, up-to-date and accurate record of their CPD activities". The guidance under that standard adds: "Your record needs to be a true reflection of the activities that you have carried out."

Standard 5 requires registrants to "upon request, present a written profile (which must be their own work and supported by evidence) explaining how they have met the Standards for CPD".

The HCPC is blunt about what "own work" means. Its guidance states that "a profile which was written by someone else, or included text from profiles produced by others, would not meet this standard". It goes further: if the HCPC suspects a profile is not your own work, for example because it contains plagiarised material, it would stop assessing the profile and investigate the matter under its fitness to practise process.

Now read the sentence that comes immediately after. The HCPC says: "This does not mean that you cannot ask a colleague for help and we would encourage you to discuss your CPD audit with your colleagues if you feel you need help."

That is the shape of the rule, and it is a sensible one. Getting help is expected. Substituting someone else's work for your own is not allowed. Nothing in that changes because the help came from software rather than a colleague.

What the regulators have published about AI so far

The joint regulator statement, February 2026

On 9 February 2026 the HCPC published a joint statement with four other statutory regulators on using AI in health and care professional education. The signatories are the General Optical Council, the General Osteopathic Council, the General Pharmaceutical Council, the HCPC and the Royal College of Veterinary Surgeons.

Read the scope carefully, because plenty of articles misrepresent it. The statement is addressed to education providers and to learners on approved programmes. It is not a rule about how a registered professional keeps their CPD record, and it does not mention revalidation.

It is still worth reading, because it tells you where regulator thinking is heading. Three of its principles translate directly to your own practice:

  • Accountability. "Learners, education providers and staff appropriately communicate where and how AI is being used."
  • Academic integrity. "Even when using AI, learners must still meet the requirements linked to each regulator's professional standards."
  • Confidentiality. Learners are expected to understand how to comply with data protection legislation and guidance in order to maintain patient confidentiality.

The statement also names the risks it is worried about, including "an overreliance on AI and the loss of core skills, as well as the potential for biased or misleading outputs", and links those risks to patient safety.

Strip it back and you get three ideas: be open about where you used it, stay accountable for what it produced, and do not let it near confidential information.

Where the NMC has got to

The NMC has not published a standalone rule on AI in revalidation. It is currently reviewing both the Code and revalidation. In November 2025 it published early findings from its review survey, which had passed 12,500 responses at that point, reporting that professionals wanted clearer standards on the safe and appropriate use of AI tools in nursing and midwifery.

"It's clear from the initial findings of our survey that nurses, midwives and nursing associates want us to invest in a modernised Code and revalidation guidance to account for the rapidly-evolving world of digital technology, including artificial intelligence." Prof Donna O'Boyle, Acting Executive Director of Professional Practice, NMC, November 2025

The NMC said it would consider those findings alongside its wider research before consulting in September 2026. So change is coming, and the direction appears to be guidance rather than prohibition.

Until then, the requirements that govern your five reflective accounts are the ones already in place. The accounts are about your own practice, they are your own work, and they must not contain anything that could identify a patient, service user or colleague.

Assistance versus authorship

Here is a test that works better than any rule. If a confirmer, an auditor or a fitness to practise panel asked you to talk through this reflection without your notes in front of you, could you?

If yes, the tool helped you. If no, the tool replaced you, and you have a problem that has nothing to do with technology and everything to do with honesty.

Reasonable assistance looks like:

  • asking a tool to question you about an event so you get past the blank page
  • taking what you have already said and arranging it into a model such as Gibbs
  • checking spelling, grammar and readability, especially if you are dyslexic
  • asking what a standard means in plain English, then checking that against the regulator's own page

Not defensible:

  • asking a tool to generate an account of an event and then lightly editing it
  • letting it invent feelings, learning or actions that were not yours
  • writing about a case you were not actually involved in
  • pasting identifiable patient information into any general purpose tool

The risk most people miss is confidentiality, not authorship

Authorship is the risk everyone worries about. Confidentiality is the one that gets people into trouble.

Consumer AI tools may retain what you type, and some use it to improve their models. That makes the copy and paste of a clinical record into a chat window a data protection problem in its own right, entirely separate from anything your regulator says about reflection.

Anonymise before you type, not afterwards. Strip names, dates of birth, NHS or record numbers, addresses, and the combination of a specific date with a specific location. Be careful with rare presentations and small units, where a bare clinical description can identify someone on its own. There is more detail in our guide to anonymising a reflective account.

A worked example

The difference between generated text and your own words is usually obvious once you see it side by side.

Generated: This experience was a valuable learning opportunity which enhanced my communication skills and reinforced the importance of patient-centred care going forward. Own words: I realised afterwards that I had explained the plan to the relative, who was standing in the corridor, rather than to the patient, who was awake and perfectly able to follow it. I did it because the relative was asking the questions and it was quicker. Next time I will address the patient first and ask them whether they want the relative involved.

The second is shorter, less polished and far more valuable. It contains a decision, a reason and a change. The first contains nothing that could not be written about any shift by anyone.

If you are reflecting after a difficult shift

Be careful with tools here, and be careful with yourself. Reflection means examining a decision so you can act differently next time. Rumination means going over the same distressing moments without moving anywhere, and a tool that keeps asking you to describe the scene again can push you towards the second one. If an event is still raw, write about what you would change and stop there.

If it is sitting heavily with you, your occupational health service, your professional body or trade union, and your employer's staff support service are all better places to take it than a blank page. Samaritans are available on 116 123, free, at any time.

If you are audited or asked about it

  • Keep the working, not just the finished document. Rough notes, a record of the questions you were asked and the answers you gave, or a first draft in your own handwriting all evidence that the thinking was yours.
  • Be able to describe your process in one or two sentences.
  • If you are asked directly, answer honestly. Trying to conceal it is a far bigger problem than the use itself.
  • Do not claim to be compliant with regulator AI guidance that does not exist yet. Say what you did.

Where Reflectory fits

This is exactly the problem Reflectory was built around. It interviews you about the event and produces a reflective account in your own words rather than drafting one for you, it screens identifiable details out of what you type, and every document it produces carries a disclosure of how it was made. You still have to do the thinking, which is rather the point.

Frequently asked questions

Is it against HCPC or NMC rules to use AI when writing a reflective account?

Neither regulator has published a rule banning it. The HCPC does require that a CPD profile presented at audit is your own work and that your CPD record is an accurate account of what you actually did. Using a tool to prompt, structure or proofread is consistent with that. Having a tool generate the reflection for you is not.

Do I have to declare that I used an AI tool in my reflection?

There is no published requirement for registrants to declare it. The February 2026 joint regulator statement on AI in education does set an accountability principle that learners communicate where and how AI is being used, and being open about it is the safer course. If you are asked directly at audit or in a reflective discussion, answer honestly.

Can I paste patient details into an AI tool and remove them afterwards?

No. Once identifiable information has been entered it may be stored or used by the provider, and the confidentiality breach has already happened. Anonymise before you type. Remove names, dates of birth, record numbers, addresses and any combination of date and location that could identify someone.