PLAB 1 is itself a multiple-choice paper, so the honest version of this article is narrower than its title suggests: the gap is not a different modality, it is the substance that rote bank drilling under-trains inside the single-best-answer format. Three things reliably slip through — the professional framing of Good Medical Practice, working knowledge of the NHS as a system, and the localisation of clinical decisions to current UK guidance. This article names those under-trained skills, gives each an observable behaviour, a deliberate-practice task, a feedback source and an exit standard, and is written for international graduates who already know their medicine but keep losing marks on UK-context items.
The official format map
The GMC's PLAB 1 is 180 single-best-answer questions in three hours: each item is a short scenario followed by a question with five options, one best answer, and you work at roughly one minute per item. It tests knowledge and its application at the standard of a doctor starting the second year of the Foundation Programme in the UK, and it is aligned to the same MLA content map as the UK student route. It is knowledge-and-application only; the performance skills of history-taking, examination and communication are assessed later, at PLAB 2 and the CPSA, not here.
| Feature | PLAB 1 (verify on gmc-uk.org) |
|---|---|
| Items | 180 single best answer |
| Duration | 3 hours (~1 min/item) |
| Standard | Doctor entering UK Foundation Year 2 |
| Blueprint | MLA content map (applies from Sept 2026) |
| Options | Five per item, one best answer |
| Performance skills | Not tested here (PLAB 2 / CPSA) |
The GMC's guide to the PLAB test sets out this format and offers a small set of example questions; treat those examples as the calibration gold-standard and confirm the current materials on the GMC site. A commercial bank supplies the volume the finite official set cannot.
Separate knowledge from performance
A correct selected answer proves one thing: that, given five pre-written options and unlimited working memory for the stem in front of you, you could recognise the best one. It does not prove that you would generate that option unprompted, that you would apply it correctly in a UK system with UK thresholds, or that you understand why the other four are wrong. On PLAB 1 the format hides this gap, because recognition is often enough to score. The gap surfaces later — and it surfaces early in the items that turn on UK-specific professional and system reasoning, where the "obviously correct" home-country answer is a distractor.
That is the real content of "what MCQ banks cannot prepare you for": not a separate skill sitting outside the exam, but the reasoning that generic disease-pattern drilling skips because it optimises for recognising the medicine rather than localising the decision.
The three under-trained skills
Good Medical Practice framing. GMC ethics and professionalism items are not general morality questions; they turn on the specific duties in Good Medical Practice — consent and capacity, confidentiality and its limits, raising concerns, probity, safe prescribing and the duty of candour. Observable behaviour: given an ethical scenario, you can name the governing GMC duty and the correct first action, not merely the intuitively "nice" one. Deliberate-practice task: for every ethics item you meet, write the single GMC principle it tests and the reason each distractor breaches or misapplies it. Feedback source: the GMC's own guidance is the rubric; an AI tutor can surface it, but the guidance is the authority. Exit standard: on unseen ethics items you can state the governing duty before you read the options, and your unseen accuracy in this strand is stable and high.
NHS context. Many items assume working knowledge of how UK care is organised — referral pathways, who does what across primary and secondary care, urgent versus routine thresholds, safeguarding routes, statutory notifications and screening programmes. Observable behaviour: you can route a presenting patient to the correct part of the NHS with the correct urgency. Deliberate-practice task: for each management item, sketch the actual UK pathway (who, where, how fast) rather than only the diagnosis. Feedback source: NHS and NICE/CKS pathway content, checked against the item. Exit standard: system and pathway items no longer produce errors driven by home-country assumptions.
UK guideline localisation. The same condition has UK-specific first-line management, thresholds and monitoring. The commonest IMG error is selecting a clinically reasonable answer that is not the current UK-recommended one. Observable behaviour: for a common presentation you can state the UK first-line step and threshold and distinguish it from the practice you trained in. Deliberate-practice task: keep a running "localisation ledger" of conditions where UK guidance differs from your prior training, each entry re-derived from current UK guidance (NICE, CKS, SIGN, the SmPC/eMC and NHS content — never a foreign formulary). Feedback source: those primary UK sources. Exit standard: on unseen management items your errors are no longer explained by non-UK guideline choices.
A four-week modality ladder
Even inside an MCQ exam, you can train these skills through escalating fidelity rather than by grinding random questions.
- Week 1 — isolated skill. Drill one strand at a time: a set of pure ethics items, then a set of NHS-pathway items, then a set of localisation items. Low speed, high analysis. Goal: build the reflex to name the governing rule.
- Week 2 — coached case. Work mixed items but pause on each to reason aloud or in writing, checking your logic against the source (guideline, GMC duty, pathway) and, where useful, an AI tutor as a fast lookup. Goal: connect the strands to real decisions.
- Week 3 — timed integrated case. Timed mixed blocks at exam pace with no assistance, so the strands compete for the same 60 seconds as everything else. Goal: transfer under time pressure.
- Week 4 — unseen simulation. Full-length, unseen, timed papers you have never worked, followed by structured error analysis sorted by strand. Goal: a clean readiness signal, not a rehearsed one.
When AI feedback helps, when it does not, and when a human is required
AI feedback is genuinely useful for fast retrieval — surfacing the relevant guideline, restating a GMC duty, explaining why a distractor is wrong — and iatroX's clinical AI is built citation-first for exactly this verification job. It is unreliable as the final arbiter of a UK-specific management threshold or an ethics judgement, because a fluent answer is not necessarily the current UK-correct one; always confirm against the primary source. And a human — a supervising clinician or examiner — is required for anything that is genuinely a performance skill, which for the licensing pathway means PLAB 2 and the CPSA, not PLAB 1. Do not use an AI score as a pass prediction. Our guides to calibrating automated feedback and auditing an AI tutor set out how to check a model before you trust it.
A concrete failure mode shows why this matters. Ask a general model for the first-line management of a common condition and it may return the option that is first-line in North American or in your home-country guidance, stated with complete confidence, because fluency is not jurisdiction-awareness. On PLAB 1 that is precisely the distractor the item is built to catch. Use the model to find and quote the relevant UK guideline, then read the guideline yourself; treat any management threshold the model gives without a UK citation as unverified until you have checked it against a primary UK source. The discipline is to let the AI accelerate retrieval while the source, not the model, sets the answer.
A balanced case and task matrix
The trap is practising only the scenarios you already find comfortable. Build a simple matrix so your practice is representative: rows for the three strands (GMC framing, NHS context, localisation) and columns for common high-yield contexts (acute medicine, primary care, paediatrics, mental health, prescribing safety, safeguarding). Track attempts and unseen accuracy in each cell, and deliberately practise the empty cells. This is the same discipline as a blueprint-coverage matrix, applied to the strands that generic banks under-sample.
| Strand \ context | Acute | Primary care | Paediatrics | Mental health | Prescribing |
|---|---|---|---|---|---|
| GMC framing | |||||
| NHS context | |||||
| Localisation |
Fill each cell with attempts and unseen accuracy; the empty and low-scoring cells are your next sessions.
A worked example: one candidate, three strands
Dr Adeyemi is an international graduate sitting PLAB 1 in five weeks. His generic-bank first-attempt accuracy is a healthy 74%, so on a headline number he looks ready. An error sort tells a sharper story: two-thirds of his wrong answers are not knowledge gaps at all. They cluster in three places — he picks the clinically reasonable option rather than the GMC-correct first action on consent and confidentiality items; he routes urgent presentations to the wrong part of the NHS; and he chooses a first-line drug or threshold from the practice he trained in rather than current UK guidance. His disease knowledge is fine; his localisation is not, and no amount of extra disease questions will fix that.
His plan follows the ladder, not more volume. In week one he drills the three strands in isolation and starts a localisation ledger, writing the UK first-line step and threshold for every management item he misses. In week two he works mixed items slowly, naming the governing GMC duty or guideline before he reads the options and checking each against the source. In weeks three and four he moves to timed, unseen, mixed blocks with no assistance and sorts every error by strand. His exit standard is explicit and measurable: on unseen items his UK-context error rate is halved, he can state the governing duty or UK threshold before seeing the options, and his unseen trend is rising across three consecutive blocks. He does not add a second bank, and he does not chase a predicted score — the strand-sorted unseen error rate is his readiness signal.
Red flags your practice is not transferring
- Memorised scripts. You recognise the answer because you have seen the exact item, not because you can derive it.
- Repeated cases. Your rising percentage is re-exposure to the same questions rather than new coverage.
- Generic feedback. Explanations that restate the answer without naming the governing UK rule or GMC duty.
- Uncalibrated scoring. Any predicted-pass number not validated against unseen, timed performance.
- No official-rubric check. You have never checked a sample of your reasoning against the GMC's own materials or current UK guidance.
Any two of these together mean your practice is training recognition, not transfer.
Bottom line
PLAB 1 is an MCQ exam, so nothing here replaces a good bank — but a bank alone under-trains the professional framing, system knowledge and UK localisation the paper genuinely tests, and it cannot touch the performance skills assessed later at PLAB 2 and the CPSA. Train the three strands deliberately, measure them on unseen timed blocks, and check your reasoning against GMC and UK sources rather than against a confident explanation. Use iatroX for the unseen-measurement job and its citation-first AI for fast verification; use a supervised simulator and real feedback for the performance skills it does not claim to cover.
Frequently asked questions
How do I know whether I have covered the full PLAB 1 blueprint? Build a coverage matrix from the MLA content map — areas of clinical practice, key presentations and conditions — and record attempted volume and unseen accuracy in each area, including the three under-trained strands above. Coverage is a property of your attempt distribution, not of a bank's headline total; an area with few attempts or low unseen accuracy is uncovered no matter how high your overall percentage.
Can one question bank be enough for PLAB 1? For the knowledge-and-application content of PLAB 1, one solid bank worked to genuine coverage, plus the GMC's official example material for calibration, is enough for many candidates. It is not enough for the professional-framing and localisation strands unless you deliberately analyse those items against UK sources, and it is not enough for the licensing pathway as a whole, because PLAB 2 and the CPSA test performance skills no bank reproduces.
What should I measure instead of my overall Q-bank percentage for PLAB 1? Measure unseen, timed, mixed accuracy and its trend, broken down by content-map area and by the three strands. A revised-bank percentage is inflated by recognition; the unseen trend is the honest proxy for exam-day performance. Our article on why a Q-bank percentage is not your exam score explains why the two diverge and how to read the gap.
When should I stop doing new PLAB 1 questions? When new questions stop surfacing new gaps. If a fresh unseen block reveals no new content-map or strand weakness and your unseen trend has plateaued at a safe level, additional new volume is low-value, and your remaining time is better spent on full-length timed papers and targeted localisation work. Volume for its own sake is not readiness.
Which PLAB 1 resource should I use for my weakest component? Match the tool to the diagnosed weakness: for GMC framing, the GMC's own guidance plus targeted ethics sets; for NHS context, NHS and NICE/CKS pathway content; for localisation, a UK-guideline ledger built from primary UK sources; for pacing, timed mixed unseen blocks; and for the performance skills, a supervised PLAB 2/CPSA simulation with real feedback. Diagnose the component with an error sort first, then choose the resource that trains that specific gap rather than adding generic volume.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026. Figures for third-party resources are vendor-reported and change without notice — confirm current details on the relevant product pages. Disclosure: iatroX operates a PLAB 1 / UKMLA question bank and competes with other banks; this article confines iatroX to jobs it can honestly claim — unseen MCQ measurement and citation-first verification — and states plainly that it does not replace a PLAB 2 or CPSA simulator or a supervising clinician for performance skills. Corrections are welcome via the feedback route on iatrox.com.
References: GMC, "A guide to the PLAB test" and the Medical Licensing Assessment / MLA content map (gmc-uk.org); GMC, Good Medical Practice (gmc-uk.org); NICE, CKS, SIGN, SmPC/eMC and NHS content for UK guideline localisation; iatroX, "Your Q-Bank Percentage Is Not Your Exam Score" (https://www.iatrox.com/blog/qbank-percentage-not-your-exam-score); iatroX, "How to calibrate automated feedback" (https://www.iatrox.com/blog/ai-graded-saqs-and-osces-how-to-calibrate-automated-feedback-before-you-trust-the-score); iatroX PLAB 1 bank (https://www.iatrox.com/plab-1).