The iatroX Socratic Tutor is a Pro feature that sits on the Q-bank. When you get a question wrong, it does not hand you a fluent explanation and move on. It asks you first. It diagnoses your specific misconception before it teaches anything. It withholds the answer to force you to retrieve what you know — because the evidence says that is how durable learning works, and handing you an explanation is how it quietly does not.
It is grounded in NICE, CKS, SmPC/eMC, SIGN, and the exam question's own official explanation and reference. It is calibrated per exam — MRCGP AKT, GPhC CRA, MRCP Part 1, DTM&H, Italian SSM. It is not ChatGPT with a medical skin. It is a tutor built on the specific question you just got wrong, using the specific sources the exam expects you to know.
The Problem with How Everyone Uses AI to Revise
The loop is familiar to every trainee who has ever pasted a question into ChatGPT: get a question wrong → copy the question stem into the chatbot → receive a fluent, well-structured, 500-word explanation → read it → feel smarter → move on → encounter a similar concept next week → go blank.
The explanation was clear. The understanding was real. The learning was not.
This is not a hypothetical risk. The Bastani et al. study, published in PNAS in 2025, showed that students given unrestricted GPT-4 access performed 17% worse on a subsequent no-AI exam than students who never had AI at all. The guardrailed "tutor" version — which gave hints rather than answers — avoided the harm entirely. The full evidence is here. The short version: the AI does the cognitive work for you, which feels productive but removes the effortful retrieval that creates durable memory.
What the iatroX Tutor Does Instead
It binds to the exact question you just got wrong. This is not a generic chatbot conversation about a medical topic. The tutor has the question stem, your answer choice, the correct answer, the official explanation, and the guideline reference as context. It knows what you got wrong and has the authoritative answer in front of it. Its job is not to tell you — its job is to help you get there yourself.
Before explaining anything, it diagnoses your specific gap. Not "you got cardiology wrong" — that is a topic, not a diagnosis. Rather: "you applied the wrong threshold," or "you confused two drug mechanisms," or "you misinterpreted the clinical scenario." The tutor identifies the specific reasoning error you made on this specific question — because correcting a specific misconception is far more effective than receiving a generic topic explanation.
Then it asks you to reason through it. Not a rhetorical question. A genuine retrieval demand. You have to articulate what you think the answer should be and why. The discomfort of not knowing — the blank, the hesitation, the "I think it's maybe..." — is the point. That is the retrieval event that strengthens the memory trace. The cognitive effort that ChatGPT removes is the cognitive effort that the iatroX tutor preserves.
Three Things That Make It Different
Socratic-first by default. It asks before it tells. Every exchange is a retrieval event. The default is question-first because the evidence says question-first produces durable learning, and explanation-first produces the illusion of learning. This is not a preference or a philosophy — it is what the experimental data shows.
Grounded in sources. Every clinical claim is anchored to NICE, CKS, SmPC/eMC, SIGN, NHS content, or the exam's own official explanation — not generated from training patterns. When the tutor says "the first-line treatment is X," there is a specific guideline or SmPC entry behind it. When it says "the threshold for this action is Y," that comes from a published source, not a pattern match across internet text. This grounding matters for accuracy, for verifiability, and for building the kind of source-aware clinical reasoning that good practice requires.
Per-exam calibration. The tutor speaks MRCGP AKT differently from GPhC CRA calculations, differently from DTM&H tropical differentials, differently from MRCP Part 1 clinical science, differently from Italian SSM. The register adapts (primary care pragmatism vs pharmaceutical precision vs tropical medicine geography). The depth adapts (GP-level management vs specialist-level pathophysiology). The clinical examples adapt (UK general practice scenarios vs tropical case presentations). The pedagogical approach adapts (guideline-application reasoning vs numerical calculation methodology vs differential-diagnosis construction).
For Exam Crunch: The "Just Explain It" Override
There are legitimate moments when you want the answer fast. The night before the exam. A concept you have revised five times and just need to confirm quickly. A calculation method you understand but want to check once more. The override exists — you can switch to direct explanation mode and get the answer without the Socratic exchange.
It is not the default, by design. The default is question-first because that is what the evidence supports for durable learning during the study phase. The override is an escape valve for time-pressure — a conscious, deliberate switch that the trainee makes when the circumstances warrant it, not the path of least resistance that the tool defaults to.
Where It Lives and What It Covers
The Socratic Tutor lives inside the iatroX Pro Q-bank — accessible on web and mobile (React Native app for iOS and Android). It covers every exam in the iatroX catalogue.
Free UK core exams: PLAB 1, UKMLA, MRCGP AKT, MRCP Part 1, MRCEM, PSA — accessible at no cost. The Q-bank questions are free; the Socratic Tutor is a Pro feature that adds diagnostic, question-first tutoring on top of the free questions.
Premium specialist and international exams: SCEs (13+ specialties), specialist diplomas (FFICM, DipIMC, DTM&H, DFSRH, DRCOG, DGM), MRCPCH (FOP/TAS/AKP), MRCPsych (A/B), FRCA (Primary/Final), GPhC CRA, dental exams (ORE Part 1, MFDS Part 1, NDEB AFK), US boards (USMLE Step 2 CK, Step 3, ABFM, ABIM, ABEM), Canadian exams (MCCQE Part I, CCFP, RCPSC), Australian exams (AMC CAT, RACGP AKT, RACP, ACEM), and Italian SSM. £29/month or £99/year — one subscription covering all exams.
The tutor connects to the adaptive engine: concepts you struggled with during the Socratic session feed back into spaced repetition scheduling, resurfacing at optimal intervals for retrieval practice. Anki export is available at /anki for trainees who want self-directed spacing alongside the platform's adaptive scheduling.