For most of a century, medical education after qualification happened in separate rooms: the study day, the conference, the e-learning module completed at year end. That separation is dissolving. The newest pattern in clinical AI is education generated from practice itself, credits and learning records built out of the real questions a clinician asked during real patient care. It is the most interesting structural change in continuing education in decades, and it deserves both enthusiasm and scrutiny.
What just-in-time education looks like
The mechanics are converging across platforms. OpenEvidence, in the US, launched free CME in April 2025: physicians revisit the clinical questions they asked, complete short assessments on them, and receive AMA credit transcripts, education literally derived from their consultation history. UpToDate Expert AI began awarding CME credit inside the answer workflow in March 2026, so the lookup and the learning record happen in one motion. In the UK, GPnotebook's AI Answers automatically logs 0.05 CPD credits for each answer a Pro subscriber reads, an approach we examined closely in our analysis of automatic CPD. Different regulatory environments, same design instinct: capture the learning where it actually occurs, at the point of care.
Why this is genuinely good
The old model had a truthfulness problem. The learning that most changes practice has always been case-driven: the unfamiliar presentation, the drug you had to look up, the threshold you checked twice. Almost none of it was captured, while the capturable learning, modules and lectures, was often the least connected to practice. Point-of-care education inverts that. It documents the questions a clinician really had, timestamps genuine curiosity, and removes the year-end scramble to reconstruct a learning portfolio from memory. For time-poor clinicians, converting work you already did into evidence you already earned is an unambiguous win.
The line to watch: exposure versus learning
The scrutiny matters too. Reading an answer is exposure; learning is what happens when you retrieve, reflect and change something. A CPD system that credits every answer read risks certifying exposure at industrial scale, and appraisal portfolios full of auto-logged fragments may satisfy a number while saying little about development. The UK's appraisal philosophy, for all its friction, is built on reflection: what did you learn, and what will you do differently? The best point-of-care education keeps that question in the loop. The worst automates it away.
What counts at appraisal
For UK clinicians the practical question is how point-of-care capture meets the appraisal room. The revalidation framework asks for CPD that is relevant to your scope of practice and, crucially, reflected upon: what you learned and what changes as a result. A raw feed of answers read demonstrates activity, and activity has some evidential value, but appraisers are entitled to ask what any of it changed. The strong pattern is curation plus reflection: let automatic capture do the remembering, then periodically promote the items that mattered, the question that changed a prescription, the threshold you had wrong, into short reflective entries written in your own words. Five reflected items outweigh five hundred logged ones, and the combination, comprehensive capture underneath, selective reflection on top, is both less work than the year-end reconstruction ritual and more honest than either extreme alone.
AI as the teacher, not just the register
The more ambitious version of this trend is coming into view: AI not merely logging education but conducting it. A system that notices you have asked about the same drug class three times this month can offer a structured refresher. One that knows which guidelines you consult can quiz you on the parts you never open. The platforms already experimenting with tutor modes and study copilots are building toward exactly this: continuous education that is personalised by your actual practice, delivered in the gaps of the working day.
Designing credit that means something
If point-of-care education is to keep its promise, the credit mechanics need designing rather than merely automating. Credit that follows assessment means more than credit that follows exposure: a short check that you can answer the question you looked up is a low-friction way to make a logged item evidence of learning rather than of reading. Credit should resist double counting, so the same glance at the same topic does not accumulate indefinitely. Reflection should be sampled rather than demanded universally, keeping the burden proportionate while preserving the habit that gives UK appraisal its meaning. And the record should remain the clinician's own, portable across employers and platforms, because a learning history locked inside one vendor is leverage, not development. Regulators and royal colleges will eventually write rules here; the platforms that pre-empt them with honest design will have earned the trust the category needs.
How iatroX handles the balance
iatroX sits deliberately on the reflective side of this shift. Clinical questions asked through the platform can be logged to My CPD with the clinician's own reflection attached, producing appraisal-ready records that document learning rather than merely exposure, and the same account connects to question banks and a Socratic Tutor when a gap deserves proper study rather than a note. Continuous education inside AI is the right destination; keeping the clinician's thinking inside it is the part worth defending.