Two of the most consequential companies in clinical AI are converging on medical learning from opposite directions. OpenEvidence, the US answer engine, is adding education: CME from real questions, society teaching content, evidence grading. AMBOSS, the education platform, has added answering: a clinical AI mode beside its study one. Watching them meet in the middle tells you what the future of medical learning looks like, and the answer is not a single winner. It is a stack.
Two directions of travel
OpenEvidence began as point-of-care literature synthesis and became the most widely used medical AI in American practice; its education layer, free CME since April 2025, educational content from societies such as ACEP, Cochrane reviews in the workflow, and the GRADE-based EvidenceGrade from July 2026, retrofits learning onto lookup. UK readers should note it has been unavailable here since its April 2026 withdrawal. AMBOSS travelled the reverse road: born in medical education, with a structured knowledge library and question banks used by students worldwide, it launched AI Mode Clinical Care in late 2025, topping the independent NOHARM safety benchmark of 31 systems, while running AI Mode Learning as a dedicated study copilot. One company teaches its searchers; the other lets its students search.
Not which is better, but which job
Head-to-head framing misleads here, because the products, and the habits they serve, occupy different layers of how a clinician actually develops. A more honest map assigns tools to jobs:
| Need | Best approach |
|---|---|
| Quick answer now | AI search over trusted sources |
| Deep understanding of a topic | Tutoring and guided explanation |
| Committing knowledge to memory | Question bank with forced retrieval |
| Keeping it long term | Spaced repetition scheduling |
| Exam readiness | A mixed strategy across all of the above |
Every row is real work, and no single interface does every row well. The future of medical learning is clinicians assembling this stack deliberately rather than hoping one subscription covers it.
What each company's move validates
OpenEvidence adding education validates the top of the stack meeting the bottom: the questions you ask in practice are the truest syllabus you have, and a platform that converts them into structured learning is closing a loop medicine always left open. AMBOSS adding clinical answers validates the reverse: knowledge built for exams should remain usable at the bedside, in the same system, with the same trusted content underneath. Both moves point at the same destination, a single environment where asking, understanding, memorising and retaining are connected, and both companies are executing credibly toward it from their own end.
Where the two models could converge next
Project both trajectories forward and they meet at the same product: a longitudinal learner model attached to a trusted knowledge base. The search company already knows what you ask; the education company already knows what you can answer. Merge those signals and the system knows what you need before you do, scheduling study from your consultation patterns and adjusting clinical answers to your demonstrated knowledge. Society content flows into both: the same guideline that grounds an answer becomes the syllabus for the module and the source of the exam item. Credentialing is the natural end point, with platforms evidencing capability to regulators rather than merely logging exposure. Neither company is fully there, and the learner-model layer raises governance questions the profession should shape early, but the direction of both roadmaps is unmistakable, and it is the same direction.
The gap that remains for UK clinicians
For a UK reader the stack has a jurisdiction problem. OpenEvidence is unavailable. AMBOSS is excellent but US-centred in its guideline orientation, which matters exactly where UK practice is most specific: thresholds, referral pathways, prescribing norms. The UK layer of the stack, answers grounded in NICE, CKS and national sources, plus exam preparation mapped to UK blueprints, has to come from somewhere built for it, a landscape we survey in our guide to AI clinical search tools for UK GPs.
What UK educators can take from both
For UK medical educators the two roadmaps read as a free strategic briefing. From the OpenEvidence direction: clinicians' real questions are the richest needs assessment available, and UK education providers should be asking how consultation-driven curiosity could feed curricula and CPD design here, within UK governance. From the AMBOSS direction: content built for learning retains its value at the point of care only if it is structured, current and trusted, which is an argument for investing in maintained UK knowledge assets rather than static lecture archives. And from both: the learner model is coming, so the bodies that own UK assessment and curricula should engage with how longitudinal capability data gets used before commercial defaults decide it for them.
The stack, assembled in one place
iatroX is the UK-grounded version of the converged model both giants are building toward: Ask iatroX for the quick-answer row, cited to UK national guidance; the Socratic Tutor for deep understanding, asking before it answers; adaptive Q-banks for memory, mapped to UK and international exam blueprints; spaced repetition for retention, scheduled automatically; and all of it in one account so the stack assembles itself. The future of medical learning is layered. Choose each layer on purpose.