Using AI for Reflective Practice: The Rules for UK Doctors

Plenty of doctors already open a general-purpose AI tool to help with an appraisal reflection. The question is not whether it happens — it does — but whether it is allowed, and where the line sits.

The short version: you can use AI to help you reflect. You cannot use it to reflect for you. The insight has to be yours, patient-identifiable data has to stay out, and you are accountable for whatever you submit. This post sets out what the regulators say, the three lines not to cross, and a workflow that keeps AI on the right side of them.

We summarise the guidance; we do not replace it. Check the current GMC position and your own college's rules before you rely on anything here — this area is moving quickly.

What the regulators say

No UK regulator bans the use of AI in reflective practice. None mandates it either. The common thread is narrower, and more useful, than a yes or no.

The GMC does not prescribe how you reflect. Its reflective-practice guidance is clear that reflection must be meaningful and demonstrate insight and learning — not a mechanical record of events. And the accountability principle at the heart of Good Medical Practice (2024) carries straight across to reflection: you are responsible for your professional record, so whatever you write, you own.

The Royal Colleges have gone further on AI specifically. The RCPCH's guidance on AI in portfolios and assessments puts it plainly: it is "neither possible nor desirable to mandate that doctors should never use AI" — but AI should supplement, not replace, reflective writing. Entries "must relate to real events and cases." It warns against a "purely mechanistic 'cut and paste' approach," and states that feedback and self-reflection "should not be generated by AI," because the value is in the authenticity.

Put together, the position is consistent: AI can support the process of reflecting. It cannot supply the substance. The learning has to be real, and it has to be yours.

The three lines you don't cross

1. Don't let AI write the reflection for you. A reflection is evidence that you thought about something and changed how you practise. An AI-generated paragraph is neither. Submitting AI-written reflection as though it were your own is treated as a probity matter, not a question of style.

2. Keep patient-identifiable information out of AI tools. This is a confidentiality duty and a data-protection one at once. Many general-purpose AI tools retain and train on what you type. Anonymise fully before any clinical detail goes near one — the same standard you already apply to written reflection.

3. You are accountable for accuracy. General-purpose AI can state things confidently and wrongly. If a tool helps you summarise a paper or a guideline, the responsibility for what you record — and then act on — is still yours. Check anything factual against the source.

Where AI genuinely helps

Within those lines, there is real and legitimate use:

• Prompting your thinking. "What might be worth reflecting on after reading this paper?" is a fair way to get unstuck — as long as the answer you write is your own.
• Structuring a draft you have already thought through. You supply the insight; the tool helps you order it.
• Summarising an article so you can spend your limited time on the reflection rather than the reading.

The distinction in every case is the same: AI does the scaffolding, you do the reflecting.

A workflow that stays inside the rules

1. Read the paper or guideline and form your own one-line takeaway — what, if anything, changes for your practice.
2. If it helps, use AI to prompt or tidy — never to invent the takeaway.
3. Write the reflection in your own words, tied to your real, anonymised cases and decisions.
4. Keep all patient-identifiable detail out of any tool.
5. Own the final entry. It should read like you, because it is you.

How the MCB CPD Tracker fits

We built the MCB CPD Tracker around exactly these lines. Its optional "Generate AI draft" feature sends only the article's details — title, journal, category, summary — to the AI. It does not send your name, your patients' data, or your own words. And it will not let a draft stand as your reflection: you have to review and personalise it before it is saved, because reflection has to be personal to count. It is AI-assisted reflection that keeps the insight, and the accountability, with you.

For the reflection itself, our CPD reflection template and the guide to logging journal reading for GMC revalidation cover the wider workflow.

Frequently Asked Questions

Can I use ChatGPT to write my appraisal reflection?

You can use it to help — to prompt your thinking or tidy a draft — but not to generate the reflection itself. The entry has to reflect your own insight and learning, and you are accountable for it. Submitting AI-written text as your own reflection is a probity issue. And never paste patient-identifiable information into a general-purpose tool.

Is using AI for reflection against GMC rules?

No UK regulator bans it. The GMC does not prescribe how you reflect, only that reflection is meaningful and shows insight; the Royal Colleges accept that AI will be used but require that it supplements rather than replaces your own reflection. Check the current GMC and college guidance, as the position is developing.

Is it safe to put patient details into an AI tool?

No. Treat it as you would any confidentiality decision and anonymise fully first. Many AI tools retain and train on what you enter, so identifiable clinical detail should never go in.

Will my appraiser mind that I used AI?

There is no rule against AI assistance. What an appraiser looks for is a reflection that shows genuine insight and a change in practice. If the entry does that and is truly yours, the tool you used to structure it is not the issue. A generic, insight-free entry is the problem — whoever, or whatever, wrote it.