This is for candidates using Pastest's MRCPCH FOP/TAS bank who want the AI "Tutor mode" to deepen the Theory and Science (TAS) paper rather than quietly hand them answers. The limitation to plan around: an on-demand AI assistant can leak the answer before you retrieve it, drift on jurisdiction, and occasionally state something a guideline does not — so the workflow below makes you commit first, prompt deliberately, and verify every material claim against official sources.
What Pastest offers for MRCPCH TAS right now
Checked on 19 July 2026, Pastest sells MRCPCH revision as an FOP/TAS package (with individual bundles also available), so your TAS practice sits inside a combined bank. Every figure here is vendor-reported; confirm it on pastest.com before purchase.
| Feature | What Pastest states (vendor-reported, 19 July 2026) |
|---|---|
| Coverage | MRCPCH FOP and TAS (combined product; TAS-relevant science and evidence-based-practice topics are a subset you filter to) |
| Question volume | Nearly 5,000 exam-style questions across FOP/TAS |
| Learning content | Over 600 textbook topics, with 3D imagery built into the textbook |
| Mocks | 10 past papers |
| AI | "Tutor mode" — instant, on-demand assistance while you answer questions |
| Analytics | Real-time performance tracking and exam-readiness insights |
| Price | 3 months £94.99, 6 months £154.99, 12 months £184.99; 48-hour free trial (limited) |
The strength is a large bank, integrated notes and an AI assistant in one place. The limitation: "Tutor mode" is a generative assistant, and a bank cannot certify readiness by itself. Both are managed by the workflow below.
The TAS exam anchor
Theory and Science is one of the first written stages of MRCPCH, frequently sat the same day as Foundation of Practice (FOP morning, TAS afternoon). RCPCH sets it as 100 single-best-answer questions in two hours, one paper, computer-based, one mark per correct answer, no negative marking. TAS is the science paper: basic scientific, physiological and pharmacological principles plus evidence-based practice. That emphasis — mechanisms, physiology, pharmacology, statistics and critical appraisal — is exactly what a prompt-driven workflow deepens. RCPCH publishes the content areas and official example questions but no fixed per-domain count, so calibrate against the official examples and treat third-party weightings as estimates.
A prompt taxonomy for the science paper
Structured prompts beat "explain this" because they target a specific reasoning failure. Six families cover almost every TAS miss: mechanism, discriminating features, option elimination, guideline and evidence verification, counterfactuals, and retrieval testing. The workflow uses all six — but only after you have committed an answer.
Twenty-five reusable prompts, grouped by error type
These are templates. Do not paste proprietary Pastest question text into any external tool; keep the question in Pastest and use these to interrogate the reasoning.
Mechanism gaps:
- "Explain the mechanism that makes the correct answer true here, at the level of physiology or pharmacology, in no more than five sentences."
- "Which single step in the pathway is being tested, and what happens clinically if that step fails?"
- "Give the mechanism for the correct answer and for the most tempting wrong answer, side by side."
- "What is the decisive or rate-limiting factor that determines the answer in this stem?"
Discriminating-feature confusion: 5. "List the two or three features that discriminate the correct condition from the one I chose, most specific first." 6. "If you changed one clinical detail in the stem, which answer would become correct, and why?" 7. "What single investigation or finding most discriminates this diagnosis, and what result would rule it out?" 8. "Tabulate my chosen option against the correct option on onset, key marker and first-line management."
Option-elimination failures: 9. "For each of the five options, give the one reason it is right or wrong — one line each." 10. "Why is my chosen option wrong specifically, not merely less right?" 11. "Which distractor is the intended trap here, and what misconception does it target?" 12. "Rank the options from most to least defensible and justify the ordering."
Guideline and evidence verification: 13. "Which UK guideline or evidence source underpins this answer, and what does it actually recommend?" 14. "Is this a NICE, CKS, SIGN, RCPCH or SmPC-based answer, and has that guidance changed recently?" 15. "For this therapeutics item, what does the SmPC on the eMC say about the relevant indication, dose basis or monitoring?" 16. "Summarise the strength of the evidence behind this recommendation in one sentence."
Counterfactual and edge-case reasoning: 17. "What single change to age, comorbidity or timing would change the correct answer, and to what?" 18. "In which patient would this answer be unsafe or wrong?" 19. "Give one clinically important exception to the rule this question teaches." 20. "What is the boundary condition where first-line management changes?"
Retrieval and consolidation: 21. "Write one new, unseen question testing the same principle, without reusing this stem." (Then close the tutor and answer it later.) 22. "Give me a three-word memory hook for the discriminating feature, no explanation." 23. "State the single rule I should be able to recall in a week from this question." 24. "Turn my one-line misconception into a corrected rule I can quiz myself on tomorrow." 25. "List the three highest-yield facts from this topic I am most likely to forget."
The anti-answer-leak rule
Before you open Tutor mode on a missed question, write down your chosen letter and a one-sentence rationale. A generative assistant's first line usually reveals the answer, and once seen you cannot un-see it — the retrieval benefit is gone. Commit first, then prompt; if you did not know it, that is data, so log it. This rule most determines whether the tutor helps or quietly inflates your sense of mastery, and it mirrors the framework pillar on auditing an AI tutor.
The jurisdiction-and-date check
TAS is a UK paediatric exam, so add one reusable prompt to every guideline-related query: "State whether this answer reflects current UK paediatric practice — RCPCH, NICE, CKS, SIGN or the SmPC — and give the date of the guidance you are relying on. If it is not UK-specific, say so plainly." Generative tools drift toward US or undated content, and a mechanistically elegant answer can still be the wrong first-line for a UK child; forcing a jurisdiction and date catches that before it becomes a memorised error. Where the tutor gives a score or confidence statement, read it through calibrating automated feedback before trusting it.
Keep a misconception record
For each meaningful miss, log four fields and nothing more: the incorrect rule you were using (one sentence), the corrected rule (one sentence), one transfer question that tests the corrected rule in a different context, and a review date. This turns a scatter of wrong answers into a finite, revisable list. The transfer question is the important part — it forces you to apply the corrected rule rather than recognise the original stem.
Run a weekly verification sample
Once a week, take five random tutor outputs and check them against a primary or official source — NICE, CKS, SIGN, the SmPC on the eMC, RCPCH guidance or the primary literature — and log any discrepancy. If the tutor is materially wrong more than occasionally on facts, use it only for reasoning scaffolds and rely on primary sources for facts. It is a small audit, but it is the difference between using an AI assistant and trusting it blindly.
Worked example: a seven-day plan around clinical work
Assume a trainee working full time, five weeks from TAS, weak on pharmacology mechanisms and statistics.
| Day | Job (about 60 min) | Tool |
|---|---|---|
| Mon | Learn: pharmacology mechanisms, commit-then-prompt on every miss | Pastest + Tutor mode |
| Tue | Consolidate: misconception record + transfer questions | Your log |
| Wed | Learn: statistics and critical appraisal, guideline-verification prompts | Pastest + Tutor mode |
| Thu | Transfer: 25 unseen mixed science questions, no assistance, timed | iatroX |
| Fri | Simulate: 100-question timed block, 2 hours | Pastest mock |
| Sat | Verify: weekly five-output verification sample; log discrepancies | Primary sources |
| Sun | Retest: spaced recall of the week's corrected rules | iatroX / Pastest |
Pastest does one clearly defined job here — teaching and explanation through commit-then-prompt — while iatroX supplies the unseen, unassisted measurement on Thursday and Sunday. Because those items are ones you have not worked in Pastest, they measure transfer rather than repeat recognition, with no proprietary-algorithm claim on either side. This division of labour is the two-Q-bank rule applied to an AI-tutor workflow.
Decision checklist
- Continue with Pastest and Tutor mode if commit-then-prompt is raising your first-attempt accuracy on unseen questions.
- Supplement with an unseen bank (iatroX) when your Pastest percentage is high but unseen accuracy lags — that is recognition, not recall.
- Switch or reduce reliance on the tutor if your weekly verification sample keeps finding factual errors.
- Stop drilling a topic when it is right, unseen, twice, at exam pace.
Bottom line
Pastest is a strong, well-populated MRCPCH FOP/TAS platform, and its Tutor mode can genuinely deepen the science paper — but only if you commit before you prompt, force jurisdiction and dates, keep a misconception record and verify a weekly sample. Left unmanaged, the same assistant flatters your percentage. Add iatroX as the unseen-measurement layer so the number you trust comes from unseen questions, read in light of "Your Q-Bank Percentage Is Not Your Exam Score".
FAQ
Is Pastest enough for MRCPCH Theory and Science on its own? Pastest is a credible single primary resource for TAS — a large bank, integrated notes and an AI assistant — but "enough" depends on how you use it. Without a source of unseen, unassisted questions you cannot separate genuine recall from recognition of items you have already worked, so most candidates are better served pairing Pastest with an unseen-measurement layer than relying on the percentage from one bank.
Which MRCPCH Theory and Science component does Pastest not reproduce well? Two things. It cannot reproduce the "unseen" condition once you have worked through the bank, which is precisely the condition a readiness signal needs; and its Tutor mode is a generative assistant, not a primary source, so it does not reliably reproduce guideline-grade certainty without your verification. It also bundles TAS with FOP, so the science-and-evidence emphasis of TAS needs deliberate filtering.
How should I verify Pastest AI answers for MRCPCH Theory and Science? Commit your own answer and rationale before opening Tutor mode; then require the tutor to name the UK source (RCPCH, NICE, CKS, SIGN or the SmPC on the eMC) and the date of the guidance; and once a week independently check five random outputs against those primary sources, logging any discrepancy. If factual errors recur, use the tutor only for reasoning scaffolds and rely on primary sources for facts.
When should I stop using Pastest and move to mixed mocks? Move to full, timed, mixed mocks once your first-attempt accuracy on unseen science questions is stable across most content areas and your pace fits the 100-question, two-hour paper — usually the final two to three weeks. Keep at least one Pastest past paper and one unseen mixed block ring-fenced for that window rather than spending them early.
How should I combine Pastest with iatroX without duplicating practice? Give each a single job: Pastest for teaching and explanation through commit-then-prompt, iatroX for unseen, timed, unassisted measurement. Because iatroX supplies questions you have not worked in Pastest, the two do not duplicate — one builds understanding, the other measures transfer. Do not re-answer the same Pastest items in iatroX; that would defeat the point of a second, unseen bank.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; vendor figures (question counts, prices, access periods and features, including "Tutor mode") are vendor-reported and change without notice — verify each on the Pastest product page before relying on it. Disclosure: iatroX operates a UK clinical question bank that competes with Pastest; it does not publish a dedicated MRCPCH bank, and its role here is confined to unseen-question measurement and transfer practice — a job the audited product does not claim. iatroX is not a Pastest product and makes no claim to Pastest's proprietary features. Corrections are welcome via the feedback route on iatrox.com.
References:
- RCPCH, Theory examinations — structure and syllabi: https://www.rcpch.ac.uk/education-careers/examinations/theory/structure-syllabi
- Pastest MRCPCH FOP/TAS (product page): https://www.pastest.com/products/mrcpch-fop-tas
- iatroX, "Your Q-Bank Percentage Is Not Your Exam Score": https://www.iatrox.com/blog/qbank-percentage-not-your-exam-score
- iatroX, auditing an AI medical exam tutor: https://www.iatrox.com/blog/how-to-audit-an-ai-medical-exam-tutor-grounding-answer-leakage-hallucinations-and-retention
- iatroX comparison hub: https://www.iatrox.com/compare
Open a fresh MRCPCH Theory and Science question and run the verification prompt in iatroX →