This audit is for international medical graduates using MedRevisions as their main PLAB 1 bank who want to know what its adaptive analytics really reward. In short: the engine optimises your performance inside the MedRevisions pool, not your performance on the real GMC paper. Its principal limitation is that a rising dashboard number can hide unattempted blueprint areas and cannot, on its own, tell you whether you are ready for 180 unseen single-best-answer questions in three hours.
That is not a criticism of the product. It is a description of what any weakness-targeting feed does, and once you understand it you can use MedRevisions for the job it does well while measuring readiness somewhere the algorithm has never seen you.
What MedRevisions offers for PLAB 1 right now
The figures below are vendor-reported and were last checked on 19 July 2026. Prices, counts and features change between exam cycles, so confirm the current numbers on the MedRevisions site before you buy.
| Feature | Vendor-reported detail (19 July 2026) |
|---|---|
| Question pool | 5,400+ exam-style single-best-answer questions, described as aligned to the 2026 MLA content map; a smaller high-yield "Study Essential" subset of roughly 1,800 |
| Exam coverage | PLAB 1 and the UKMLA Applied Knowledge Test, treated as a shared MLA-mapped pool |
| Mock exams | 30+ full-length mocks, described as customisable with resets |
| Adaptive / AI features | "MedRevisions AI Professor" tutor, "Weakness Mocks" using spaced repetition, an "AI Study Assistant" study plan and "AI Flashcards" |
| Analytics | A "Readiness Score", plus performance tracking by Areas of Clinical Practice |
| Price | From around $10.83–$10.99 per month (vendor-reported; verify current tiers, currency and access length on the pricing page) |
| Free access | 70+ sample questions with limited tutor access, no card required |
Two things are worth flagging immediately. The pricing is quoted in US dollars, which suits an international audience but means you should check the exchanged cost and the exact access window before committing. And the AI features are vendor-named; this article does not claim to know how any of them work internally, and neither should you. Treat every proprietary label as a black box until you have tested it against unseen questions.
The exam you are actually sitting: PLAB 1 and the MLA content map
PLAB 1 is a single paper of 180 single-best-answer questions sat in three hours, set by the General Medical Council. It is built on the MLA content map — the same blueprint that underpins the UK Medical Licensing Assessment — and is pitched at the knowledge a doctor needs entering the second Foundation year. The content map is organised around areas of clinical practice, presentations and conditions rather than neat organ-system silos, and the January 2026 revision applies to sittings from September 2026.
Two features of that blueprint matter for anyone reading a bank's analytics. First, the GMC does not publish a public question-by-question weighting you can reverse-engineer; the official sample questions show the style and standard, not the exact domain split. Second, "aligned to the MLA content map" is a vendor claim, not a GMC endorsement. No third-party bank is the exam. The most useful thing a bank can do is expose you to the breadth of presentations at roughly the right standard — and the most useful thing you can do is check that breadth yourself rather than trust a home-screen average.
Every metric on the MedRevisions dashboard, defined
Adaptive analytics are only helpful if you know what each number counts. Here is what the common metrics mean and where each one quietly misleads.
- First-attempt accuracy — the percentage of questions you got right the very first time you saw them. This is the only accuracy figure that behaves anything like an exam, because the exam is always a first attempt.
- Repeat accuracy — your score on questions you have already seen. It rises naturally as you memorise items and tells you almost nothing about readiness.
- Percentile — your rank against other users. It moves with the population's revision stage, not just your ability, and it is not a probability of passing.
- Readiness or predicted score — a vendor-modelled estimate. It is a motivational signal, not a GMC forecast; no bank can predict your PLAB 1 result.
- Coverage — how much of the pool you have attempted. High coverage of the bank is not the same as coverage of the blueprint, and the two diverge whenever the pool is uneven.
- Difficulty — usually derived from how other users performed on an item, not from any official calibration.
- Time per item — seconds spent per question. On PLAB 1 you have roughly one minute per item, so this is one of the more transferable numbers on the screen.
The single most important discipline is to read first-attempt accuracy on unseen items and treat everything else as context.
What the adaptive engine is actually optimising
A weakness-targeting feed does exactly what its name suggests: it shows you more of what you get wrong. That is pedagogically reasonable, but it has a statistical consequence. Your visible average is calculated on a diet the algorithm has deliberately skewed towards your weak spots, so it understates your true standing early and then climbs steeply as those weak spots are drilled. Neither the low nor the later high is a clean estimate of exam performance.
There is a second, subtler effect. Because the feed prioritises high-volume, easily-tagged topics where it has plenty of items, low-frequency material — ethics and consent vignettes, image-dependent questions, prescribing calculations, statistics — can be under-served without ever showing up as a red bar. The engine is optimising engagement with your errors inside the pool. It is not optimising your coverage of the blueprint, and it has no way of knowing what the GMC will actually ask. That gap between "better at this bank" and "ready for that exam" is the whole reason to keep an unseen measurement layer. Our standing caveat on this is Your Q-Bank Percentage Is Not Your Exam Score.
Run your own blueprint audit
Do not trust the home-screen average to tell you about coverage. Once a week, export or read off your attempted-question distribution by area of clinical practice and compare it against the spread of presentations in the MLA content map. You are looking for areas where your attempted count is a fraction of your busiest topics — those are blind spots the feed has quietly created. The method is the same one we set out in question-bank completion is not coverage: build a simple matrix of blueprint area against attempted count and first-attempt accuracy, and treat any low-count cell as an override target regardless of what your overall percentage says.
The readiness signal the dashboard cannot fake
A credible readiness signal has five properties, and a personalised feed removes several of them by design. To trust a number as a proxy for PLAB 1, the block must be:
- Unseen — questions you have not attempted before, so you are testing recall and reasoning, not memory of an item.
- Timed — at roughly one minute per question, because pacing failure is a common way to lose marks you actually knew.
- Mixed — a random spread across the blueprint, not a topic-filtered set, so you practise switching context the way the real paper forces you to.
- Unassisted — no explanations, hints or tutor open beside you.
- Adequately sampled — at least 100 to 150 items before you read anything into the percentage; a 20-question block is noise.
If you cannot tick all five, you have a study metric, not a readiness metric.
Override rules: what to force into the feed
Because the algorithm optimises for its own high-volume topics, you should manually force these categories at least weekly, whatever your dashboard suggests:
- Low-volume domains — palliative care, ophthalmology, ENT, dermatology and sexual health, which are easy to under-sample.
- Image-dependent items — rashes, fundoscopy, ECGs, radiographs and clinical photographs, where recognition is the skill.
- Ethics, consent, capacity and safeguarding — high-yield on the MLA map and frequently under-drilled.
- Prescribing and calculations — drug doses, infusion rates and fluid maths, using the SmPC via the electronic medicines compendium (eMC) as your reference source rather than any single formulary shorthand.
- Statistics and evidence — sensitivity, specificity, predictive values and number needed to treat.
Worked dashboard example: from analytics to next week's quotas
Suppose your MedRevisions dashboard shows an overall accuracy of 71%, a readiness score labelled "on track", 62% of the pool attempted, and first-attempt accuracy of 58% on cardiology, 54% on endocrinology and 49% on ethics — but only 40 ethics questions attempted against 300+ in cardiology. Do not read the 71% as a pass signal, and ignore the readiness label entirely.
Convert the picture into quotas instead. Next week: 40 fresh ethics and consent items (closing the low-count blind spot), 30 endocrinology, 30 cardiology first-attempts, plus one 150-question mixed, timed, unseen mock with no assistance. The mock — not the dashboard — is your readiness reading. You are turning analytics into an action list without ever pretending to know your probability of passing. Repeat weekly and watch the first-attempt line on mixed mocks, not the headline average.
A seven-day pattern for international graduates
This splits the week so MedRevisions does one defined job — high-volume content review with immediate feedback — and an unseen bank measures transfer. No proprietary-algorithm claims are made about either tool.
| Day | MedRevisions job | Unseen-measurement job (iatroX) |
|---|---|---|
| Mon | 40 questions in two weak areas, read every explanation | — |
| Tue | 40 questions, force low-volume domains | — |
| Wed | 30 image and ECG items | 20 fresh mixed items, timed, no notes |
| Thu | 40 questions, ethics and prescribing focus | — |
| Fri | 30 questions; review error log | — |
| Sat | — | 150-item mixed, timed, unseen mock |
| Sun | Re-drill only the reasoning errors from Saturday | Log first-attempt accuracy by domain |
The Saturday mixed block on a bank you do not revise from — iatroX runs free UK-core PLAB 1 content for exactly this — is your weekly readiness reading. Everything MedRevisions shows you is a study metric; the unseen block is the exam-like one. That division of labour is the two-Q-bank rule: one bank to learn on, a different one to measure on, so your calibration is never contaminated by items you have already seen.
Three mistakes this audit is designed to stop
Reading the readiness score as a pass probability. It is a vendor model built on your behaviour inside one pool, not a GMC forecast. Anchor to first-attempt accuracy on unseen mixed blocks instead.
Mistaking bank completion for blueprint coverage. Finishing 100% of an uneven pool can still leave whole presentations barely touched. Audit attempted counts by domain, not just the percentage done.
Never testing yourself cold. If every question you attempt comes with an explanation a click away and a tutor in the sidebar, you have never rehearsed the actual exam condition — unaided recall under time pressure.
Decision checklist: continue, supplement, switch or stop
- Continue if first-attempt accuracy on unseen mixed blocks is rising, your blueprint audit shows no starved domains, and your pacing sits near one minute per item.
- Supplement — add a second, unseen bank for measurement — if your MedRevisions average looks healthy but you have never scored yourself cold, or your attempted-count matrix shows blind spots.
- Switch your primary learning bank only if you find sustained factual errors or the explanations are not moving your unseen scores after a fair trial — not on novelty.
- Stop buying more questions when unseen first-attempt accuracy has plateaued across the blueprint and your limiting factor is pacing or nerves; at that point mixed timed mocks, not new items, are the intervention. Compare your options on the iatroX comparison hub.
The bottom line
MedRevisions is a credible high-volume PLAB 1 bank with a modern analytics layer, and for the job of grinding through breadth with immediate feedback it is a strong option worth including in the revision stack. What its adaptive engine optimises, however, is your performance inside its own pool. It cannot certify readiness, and its headline numbers systematically flatter you as you drill. Use it to learn, audit your own coverage against the blueprint, and keep one unseen, timed, mixed block each week as the number you actually trust.
Frequently asked questions
Is MedRevisions enough for PLAB 1 on its own? For most disciplined candidates it can be a sufficient primary bank, because its vendor-reported pool of 5,400+ MLA-mapped questions (19 July 2026) covers the breadth PLAB 1 tests. The caveat is measurement: a single bank you learn on and test on will overstate your readiness, so "enough" means enough to learn from, paired with unseen mixed blocks somewhere else to check transfer.
Which PLAB 1 component does MedRevisions not reproduce well? No question bank reproduces the pressure of a genuinely unseen, mixed, three-hour paper, and a weakness-targeting feed is the opposite of a random blueprint sample. It also tends to under-serve low-volume, image-dependent and calculation material unless you force those categories. The exam skill it least reproduces is unaided context-switching at pace — which you only rehearse in cold mixed mocks.
How many MedRevisions questions should I complete per day for PLAB 1? A sustainable target is 40 to 60 questions a day with full review of every explanation, scaling up in the final fortnight. Raw volume is not the goal; a slower 30 questions fully understood beats 100 skimmed. Reserve one longer mixed, timed block each week for measurement rather than adding more daily items.
When should I stop using MedRevisions and move to mixed mocks? Shift the balance towards mixed timed mocks once your first-attempt accuracy on unseen blocks has plateaued across the blueprint and your blueprint audit shows no starved domains. At that stage new questions add little and full-length timed papers rehearse pacing and stamina, which are usually the remaining limiting factors.
How should I combine MedRevisions with iatroX without duplicating practice? Assign each tool one job. Learn on MedRevisions — drill weak areas, read explanations, build breadth — and measure on iatroX with fresh, timed, mixed PLAB 1 blocks you never revise from. Because the two pools differ, you avoid re-testing memorised items and keep an uncontaminated readiness signal, exactly as the two-Q-bank rule intends.
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 and feature names) are vendor-reported on that date and change between exam cycles — verify the current numbers on the MedRevisions product and pricing pages. Disclosure: iatroX operates a competing PLAB 1 question bank; this article confines iatroX's role to unseen readiness measurement, a job MedRevisions' adaptive feed is not designed to perform. Corrections are welcome via the feedback route on iatrox.com.
References: General Medical Council — PLAB 1 format and the MLA content map (gmc-uk.org); MedRevisions product and pricing pages (medrevisions.com, vendor-reported); iatroX PLAB 1 bank; Your Q-Bank Percentage Is Not Your Exam Score; the two-Q-bank rule.