This review is for trainees who already hold, or are considering, Pastest's MRCPCH bank and want to know whether its AI Tutor genuinely helps with Theory and Science (TAS) — the science-and-mechanisms written paper — or whether it mainly reformats an answer they should be reasoning through themselves. The short version: Pastest covers TAS well as content, but the AI Tutor is a support layer whose source grounding is not publicly documented, so it must be interrogated, not trusted by default.
One structural point shapes everything below. Pastest does not sell a standalone TAS bank; TAS is bundled with Foundation of Practice in a combined FOP/TAS product. The AI Tutor you are evaluating therefore sits across both papers, and the TAS-specific work is a subset of that bank. That is not a problem in itself, but it means "the Pastest TAS tutor" is really "the FOP/TAS tutor answering TAS items," and you should judge its behaviour on TAS-style science questions specifically.
What Pastest offers for MRCPCH TAS right now (last checked 19 July 2026)
Figures are vendor-reported, read from pastest.com on 19 July 2026. Verify on the product page before purchasing.
| Field | Finding (19 July 2026) |
|---|---|
| TAS coverage | Sold within a combined MRCPCH FOP/TAS bank (no standalone TAS product) |
| Question count | Vendor-reported ~4,000–5,000 exam-style questions (Pastest's product page cites "almost 5,000"; its collections page cites "4,000+") plus 10 past papers — verify the live figure |
| Teaching content | "Over 600" textbook topics; recently added 3D imagery |
| AI feature | "Tutor mode" / Pastest AI Tutor, described as "instant, on-demand AI support while you answer questions" |
| Analytics | Real-time progress and readiness insights (vendor-reported) |
| Price | Vendor-reported: 3 months £94.99, 6 months £154.99, 12 months £184.99; 48-hour free trial with limited access |
| Vendor performance claim | "9% increase on the average pass mark" — a marketing claim; treat with caution |
The pass-mark-uplift figure is a vendor marketing claim, not an independent finding, and should carry no weight in your decision. The question counts differ between Pastest's own pages, which is itself a reason to confirm the live number rather than quote one.
The exam anchor: what MRCPCH TAS actually is
TAS is one of the two first-stage written exams of MRCPCH, sitting alongside Foundation of Practice; candidates may take the theory papers in any order, and TAS can be sat the same day as FOP. Its format is fixed: 100 single best answer (SBA) questions, 2 hours, one paper, one mark per question and no negative marking. The theory exams are single-best-answer throughout, the older extended-matching format having been retired.
TAS is the science paper. It tests the basic scientific, physiological and pharmacological principles that underpin clinical practice, together with evidence-based practice — statistics, study design and critical appraisal. It samples the same broad paediatric content domains as the other theory exams but with a mechanisms-first emphasis: why a drug acts as it does, how a physiological system fails, what an investigation is actually measuring. RCPCH publishes one official TAS sample paper with an answer key and reasoning; note that the FOP and TAS specimens run on an older interface and include some question styles no longer used, so treat them as a guide to content and reasoning rather than to the current on-screen format.
Because TAS rewards mechanistic understanding over pattern-matching, it is a demanding test of whether an AI tutor can explain why rather than merely assert which — which is exactly what this audit probes.
Testing methodology: the rubric to run on the tutor
You cannot judge an AI tutor from marketing copy; you judge it by running a fixed, representative set of items through it and scoring its behaviour against a published rubric. Use six item types that between them expose most failure modes: a straight-recall fact, a diagnosis stem, a next-investigation question, a management question, an ethics or evidence-based-practice item, and one deliberately ambiguous stem where two options are defensible. Apply the same rubric to each, and apply it yourself rather than taking the tutor's confidence as the score.
- Correctness: does the tutor reach the answer the official key supports?
- Grounding: does it name a source you can check, or assert without one?
- Reasoning: does it explain the mechanism, or restate the option?
- Uncertainty: on the ambiguous item, does it flag the ambiguity or force a false certainty?
- Premise-handling: if you feed it a wrong premise, does it correct you?
This is the same discipline set out in the iatroX pillar on auditing an AI medical exam tutor — grounding, answer leakage, hallucination and retention — applied to TAS science content.
Grounding audit: inspecting provenance without reproducing content
Pastest does not publicly document how its AI Tutor is grounded — which model it uses, whether its answers are constrained to Pastest's own explanations, or whether it can draw on external guidance. In the absence of that documentation, treat provenance as unverified and inspect it yourself. Ask the tutor directly where a claim comes from; if it names a guideline, check the claim against the primary source rather than accepting the paraphrase. For TAS pharmacology and therapeutics, the UK references are the SmPC via the electronic Medicines Compendium (eMC), NICE and CKS, and SIGN where relevant; for evidence-based-practice items, check the statistical claim against a standard text. You are not looking to reproduce copyrighted explanations — you are checking that a claim you are about to memorise is true and current.
A practical tell: a well-grounded answer survives the question "which source says so, and does that source still say it?" An ungrounded one dissolves into confident generality. Run that test on any answer you intend to trust.
Reasoning behaviour and exam fidelity
On reasoning, the questions are whether the tutor asks a useful diagnostic question before answering, whether it explains the mechanism or merely reveals the option, whether it handles uncertainty honestly, and whether it corrects a false premise you introduce. A tutor that leaps to the answer trains recognition, not understanding — which is the opposite of what TAS rewards.
On fidelity, test whether the tutor respects the exam's context: UK paediatric practice, TAS's science-and-mechanisms weighting, the one-mark, no-negative-marking scoring, and the roughly 72-second-per-item pace of a 100-question, 2-hour paper. A tutor that gives beautiful but sprawling answers you would never have time to read in the exam is teaching a habit the clock will punish. Fidelity means the support fits the paper you are sitting, not a generic global one.
Failure modes to watch for
- Hallucinated citations: a named guideline or figure that does not exist or does not say what is claimed.
- Overconfident wording: definitive language on a genuinely contested point.
- Outdated guidance: advice superseded by a newer NICE, CKS or SIGN position.
- Answer leakage: revealing the key before you have committed, which quietly destroys the diagnostic value of the item.
- Plausible-but-unexamined elaboration: fluent extra detail that reads well and was never checked.
Safe-use protocol: answer first, interrogate second, verify third
Use a fixed three-step routine. Answer first: commit to your SBA choice before opening the tutor, so the item still measures you. Interrogate second: ask the tutor to explain the mechanism and to name its source — high-value prompts are "why is the other option wrong?" and "what would change the answer?" rather than "what is the answer?" Verify third: for anything you intend to memorise, confirm the load-bearing claim against a primary UK source. This mirrors the calibration discipline in the iatroX pillar on calibrating automated feedback before you trust the score, and it is the same citation-first habit iatroX builds into Ask iatroX.
Worked example: a seven-day TAS week around clinical work
This plan uses Pastest for one defined job — learning TAS science with tutor support — and iatroX for unseen transfer measurement. No proprietary-algorithm claims are made about either.
| Day | Primary job (Pastest FOP/TAS) | Transfer check (iatroX, unseen) |
|---|---|---|
| Mon | 30 TAS SBAs on pharmacology; commit, then interrogate the tutor on mechanisms | — |
| Tue | 20 TAS SBAs on physiology; verify two tutor claims against primary sources | 10 unseen UK-core items, timed |
| Wed | Rest / clinical day; re-state 5 mechanisms in your own words | — |
| Thu | 30 TAS SBAs on genetics and metabolism; log any tutor over-confidence | 10 unseen items, timed, no notes |
| Fri | 20 TAS SBAs on evidence-based practice and statistics | — |
| Sat | One 100-item timed block at exam pace, tutor off | 15 unseen items on the week's themes, timed |
| Sun | Review only; re-test the 10 most-missed as fresh items | — |
Keeping the tutor off during Saturday's timed block matters: the exam has no tutor, so at least one weekly session must rehearse unsupported recall. If your Pastest percentage climbs while the unseen iatroX score stalls, you are leaning on the tutor rather than learning — the pattern described in Your Q-Bank Percentage Is Not Your Exam Score. Running a second, unseen source this way is the two-Q-bank rule in action.
Decision checklist: continue, supplement, switch or stop
- Continue Pastest if its TAS explanations teach you mechanisms and the tutor survives your grounding checks.
- Supplement with unseen iatroX blocks once your first-pass percentage clears about 70% but timed, tutor-off scores lag.
- Switch or add a second paediatric bank only if you find measurable blueprint gaps, not because a competitor is newer.
- Stop relying on the tutor when your unsupported, timed scores are stable across two weeks — at that point it is a crutch, not a teacher.
Bottom line
Pastest is a substantive TAS resource, and its AI Tutor is a reasonable support layer — provided you treat it as something to interrogate rather than obey. Its grounding is undocumented, so verify; its answers can leak, so commit first; its fidelity to the TAS clock and UK context must be checked, not assumed. Use it to learn mechanisms, use iatroX to measure whether those mechanisms transfer under unseen, timed conditions, and let the official RCPCH sample paper keep your expectations honest.
Frequently asked questions
Is Pastest enough for MRCPCH Theory and Science on its own? For many candidates its combined FOP/TAS bank supplies enough coverage and teaching to be the primary resource, but "enough" depends on whether you can convert its questions into transferable understanding rather than recall of seen items; pair it with an unseen measurement source and the official RCPCH sample paper to confirm readiness, and verify the live question count and price on pastest.com.
Which MRCPCH Theory and Science component does Pastest not reproduce well? The one thing no bank reproduces is unsupported, unseen performance: with the AI Tutor available on tap, Pastest is strong at teaching but cannot, on its own, tell you how you will reason without help, which is precisely what the exam measures — so the tutor-off, unseen check is the gap you must fill elsewhere.
How should I verify Pastest AI answers for MRCPCH Theory and Science? Ask the tutor to name its source, then check the load-bearing claim against a primary UK reference — SmPC via the eMC, NICE, CKS or SIGN for pharmacology and management, and a standard statistics text for evidence-based-practice items — and discard any confidently worded claim that no current source supports; provenance for the tutor is not publicly documented, so this verification step is not optional.
When should I stop using Pastest and move to mixed mocks? Move to full 100-item timed blocks, tutor off, once your first-pass percentage sits consistently above about 70%; the drill teaches, but only the timed, unsupported mock rehearses the 2-hour TAS paper and its roughly 72-second pace.
How should I combine Pastest with iatroX without duplicating practice? Use Pastest for coverage and mechanism teaching and iatroX only for unseen, timed transfer checks, so the two never present the same item; the point of the second source is to measure retention on questions you have not seen, not to repeat Pastest's practice or its explanations.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; platform question counts, prices, features and AI behaviour change frequently and are vendor-reported — verify current figures on the vendor's own product pages before purchasing, and note that the AI Tutor's grounding is not publicly documented.
Disclosure: iatroX operates a UK clinical question bank and Socratic Tutor that competes for revision time with the product reviewed here. iatroX does not currently publish an exam-specific MRCPCH TAS bank; its role in this article is confined to the unseen-measurement and transfer-practice job — and the citation-first verification habit — that Pastest's tutor does not claim to guarantee. Corrections are welcome via the feedback route on iatrox.com.
References:
- RCPCH — Theory exams: structure and syllabi. https://www.rcpch.ac.uk/education-careers/examinations/theory/structure-syllabi
- RCPCH — Theory exams: sample papers. https://www.rcpch.ac.uk/education-careers/examinations/theory/sample-papers
- Pastest — MRCPCH FOP/TAS. https://www.pastest.com/products/mrcpch-fop-tas
- iatroX — How to audit 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 — Calibrating automated feedback before you trust the score. https://www.iatrox.com/blog/ai-graded-saqs-and-osces-how-to-calibrate-automated-feedback-before-you-trust-the-score
- iatroX — Your Q-Bank Percentage Is Not Your Exam Score. https://www.iatrox.com/blog/qbank-percentage-not-your-exam-score
Open a missed MRCPCH Theory and Science item in the iatroX Socratic Tutor →