Pastest's Tutor mode can explain a Clinical Problem Solving item the instant you open it — which is precisely the risk. This is the implementation companion to our Pastest MSRA grounding audit: where that piece tests whether the AI is trustworthy, this one shows how to use it so it sharpens your reasoning instead of replacing it — and where, on this exam, you must not point it at all.
The commit-first rule
Write your answer and a one-line rationale before Tutor mode opens. Every time. On an exam with a brutal CPS clock, the temptation to consult first and commit later is strong and self-defeating: it converts retrieval practice into assisted reading and trains a dependence the exam does not permit. The rule is non-negotiable, and everything below assumes it.
Where the tutor helps — and where it must not go
CPS items have verifiable answers an AI can ground in guidance, so the tutor is genuinely useful there. Professional Dilemmas is different: it is scored against expert consensus about judgement, there is no guideline to cite, and a fluent AI rationalisation of a ranking is actively harmful because it teaches a confident wrong framework. So the boundary is hard: use Tutor mode for CPS reasoning; do not ask it to adjudicate PD rankings. For PD, the only authority is the official rationale. This single boundary prevents the most common way MSRA candidates misuse an AI tutor.
Reusable CPS prompts that build reasoning
Keep these generic and never paste proprietary question text. For a miss: "Without giving the answer, name the discriminating feature between my choice and the correct option." For jurisdiction: "Does this reflect current UK primary-care practice? Flag any divergence and give the source and its date." For depth control: "Answer at MSRA level, not tertiary level — what would a GP actually do next?" For pace: "State the single fastest way to reach the answer under a 50-second budget." For transfer: "Give me one unseen question on this concept in a different presentation." These push the tutor towards discrimination, jurisdiction and pace — the three things CPS actually tests — rather than towards a comfortable re-explanation.
The false-premise habit
Once a session, feed the tutor a plausible wrong claim on purpose ("since trimethoprim is first-line for UTI in pregnancy…") and confirm it corrects you rather than building on the error. This is both a reliability check and a reasoning exercise: it keeps you alert to the failure mode where an agreeable assistant confirms half-remembered rules, which is where a lot of MSRA marks quietly leak.
The misconception record
For every real CPS miss, write four lines: the wrong rule, the corrected rule, one transfer question, one review date. For PD misses, write two lines instead: the principle you misjudged, and the consensus principle from the official rationale. Different papers, different records — but both dated and both testable, which is what turns an explanation into learning.
The weekly verification sample
Take five tutor outputs at random each week and check them against official or primary sources; log discrepancies. Where you want a fast cited second opinion on a UK-guideline point, Ask iatroX returns one with the source attached. A clean log earns trust; a dirty one just protected your ranking.
A seven-day pattern for applicants
Monday: 50 Pastest CPS questions, commit-first, reasoning prompts on misses. Tuesday: a timed PD session on official material — no AI. Wednesday: a timed, unseen 50-question CPS block in iatroX (free for MSRA), pure measurement. Thursday: forced-pace CPS drilling plus the weekly verification sample. Friday: second PD session. Saturday: full both-paper simulation, same-day review. Sunday: rest. Pastest's AI sharpens CPS reasoning; official material owns PD; iatroX measures on unseen items.
Two worked exchanges: where the tutor helps and where it must stop
The boundary between "use the AI" and "keep the AI out" is easiest to see through two examples on the same evening.
A CPS miss: you chose oral antibiotics for a child with a clinical picture the key marks as requiring admission. The right exchange interrogates the threshold — "without giving the answer, what single feature in this vignette should have shifted me from community to hospital management, and what is the UK source for that threshold?" The tutor names the red flag and, ideally, the guideline behind it; you verify the source, write the corrected rule, and you have converted a wrong click into a safety-netting principle you will carry into practice as well as the exam. This is the AI at its best: a fast, specific, groundable discriminator.
A PD miss: you ranked five responses to a colleague's suspected impairment and scored below the consensus. The wrong instinct is to ask the tutor "why is my ranking wrong?" — and the wrong outcome is a fluent, confident justification of an ordering the model has no authority to adjudicate, because the scoring reflects an expert consensus the AI cannot see. The right move is to close the tutor and open the official rationale, which encodes the actual judgement standard: patient safety first, then honesty and appropriate escalation, then collegial support. You study the principle, not the AI's guess at it. Getting this boundary wrong is the single most common way candidates misuse an AI tutor on the MSRA — they let a plausible voice train them into a confident wrong framework on the paper where frameworks matter most.
Run every evening's misses through that filter: verifiable clinical answer, use the tutor and ground it; consensus-scored judgement, go to the official source. The discipline is what keeps the AI an asset rather than a liability.
A realistic week, annotated
The seven-day pattern above is not aspirational padding — each slot has a job. Monday's commit-first CPS drilling builds retrieval strength on your weak domains. Tuesday's timed PD session is the appointment most candidates skip and most regret skipping. Wednesday's unseen block is your only uncontaminated readiness reading of the week. Thursday's forced-pace work targets the clock specifically, because CPS is lost on time more than on knowledge. Friday's second PD session compounds Tuesday's. Saturday's full both-paper simulation rehearses the mental gear-change between judgement and clinical reasoning within one sitting — itself a small trainable skill. Sunday's rest protects the consolidation the week depends on. Miss a day and you know exactly which job went undone.
Continue, supplement, switch or stop
Continue while commit-first holds and verification stays clean. Supplement with unseen timed CPS blocks when the tutor feels productive but measured performance is flat. Switch only on logged failures. Stop all AI in the final fortnight's simulations.
Three mistakes this workflow is designed to stop
First, pointing the tutor at Professional Dilemmas. It is the single most common misuse on this exam, and it is actively harmful — a fluent AI justification of a judgement ranking trains a confident wrong framework against a consensus standard the model cannot see. Keep the tutor on CPS, and let the official rationale own PD. Second, consulting before committing. On a paper with a punishing clock the temptation to peek is strong, but every pre-commitment consultation converts a retrieval rep into an assisted read and quietly erodes the very performance you are trying to build. Third, skipping the verification sample. An AI tutor's reliability drifts as the platform iterates and as guidance moves; a weekly five-output audit against primary sources is the cheap insurance that keeps your trust calibrated rather than blind — and on the CPS strand, where answers track current UK practice, that currency check is where stale advice gets caught before the exam catches you.
Frequently asked questions
Is Pastest enough for MSRA on its own? It can anchor CPS, but PD needs official consensus-scored practice and readiness needs unseen timed blocks — no single platform is sufficient for both papers.
Which MSRA component does Pastest not reproduce well? Professional Dilemmas scoring — no AI tutor can adjudicate judgement rankings against a hidden consensus standard, so keep the tutor away from PD entirely.
How should I verify Pastest AI answers for MSRA? Demand a named dated source for CPS claims, check via primary sources or Ask iatroX, run a weekly five-output audit, and treat any confident PD ranking from the AI as untrustworthy by default.
When should I stop using Pastest and move to mixed mocks? When CPS coverage and pace criteria are met — the final two to three weeks belong to both-paper simulations with AI closed.
How should I combine Pastest with iatroX without duplicating practice? Pastest for CPS drilling and reasoning prompts; iatroX for free unseen adaptive CPS measurement and Socratic repair; official material for PD.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; MSRA timings are NHS England's; Pastest features vendor-published and evolving. Disclosure: iatroX offers a free competing MSRA bank and Socratic Tutor. Corrections via the feedback route on iatrox.com. References: NHS England MSRA structure (medical.hee.nhs.uk); Pastest MSRA pages (pastest.com); related reading: the Pastest MSRA grounding audit and why your Q-bank percentage is not your exam score.