This audit is for UKMLA and PLAB 1 candidates — especially international graduates — using MedRevisions as a high-volume applied-knowledge-test bank and trying to read its analytics honestly. MedRevisions covers the AKT well and reports useful strength-and-weakness data; its principal limitation is that it is a knowledge bank, so it does not reproduce the CPSA, and, like any adaptive-flavoured feed, its dashboard percentages can drift away from what a fresh unseen block would show.
Current-state box (vendor-reported, 19 July 2026)
| Item | MedRevisions (vendor-reported) |
|---|---|
| Live question count | "5,400+" UKMLA/PLAB questions, updated each exam cycle |
| Components covered | UKMLA AKT and PLAB 1 (shared MLA content map) |
| CPSA / OSCE | Not covered (knowledge bank) |
| AI / analytics | "AI Medical Professor" 24/7; "Smart Analytics" by clinical-practice area; "Study Essential Mode" (~1,800 high-yield); "Weakness Mocks" using spaced repetition on performance history |
| Access / price | From "$10.99/month" (USD), flexible plans; extensions if exam dates change |
These are vendor claims taken from the MedRevisions site on 19 July 2026; confirm the current count, price and features on the product page before relying on them.
Exam anchor
The UKMLA applied knowledge test — the MSC AKT for UK students, PLAB 1 for IMGs — is built on the GMC MLA content map, updated for the 2026 cycle and applying to all MLA assessments from September 2026. The map's areas of clinical practice, patient presentations and conditions are the official coverage target; a vendor's internal categories are a third-party approximation of it, not the blueprint itself. The separate CPSA assesses clinical and professional skills OSCE-style and sits outside any MCQ bank. Keep official requirements and vendor claims in separate columns when you audit.
Every metric, defined
- First-attempt accuracy. Your score the first time you see an item — the only percentage that behaves like a test. Track this, not the blended figure.
- Repeat accuracy. Your score on items you have seen before. It rises with exposure and measures recognition, not readiness.
- Percentile. Your rank against other users, which depends entirely on who is in the pool and when — not against the standard-setting of the real exam.
- Predicted score. A modelled estimate. Treat any predicted pass as a hypothesis, never a result, and never let it substitute for an unseen block.
- Coverage. How much of the vendor's category tree you have touched — useful, but the vendor tree is not the GMC content map.
- Difficulty. Usually a cohort-derived index (how others fared). It tells you an item is discriminating; it does not tell you it matches the exam standard.
- Time per item. The most under-used metric. Against the roughly one-minute AKT pace, it flags whether accuracy is real or bought with time you will not have.
Selection bias: why adaptive feeds distort the percentage
MedRevisions' "Weakness Mocks" and spaced-repetition features deliberately re-serve your weak and previously-missed items. That is good teaching, but it corrupts the home-screen average in two ways. First, an adaptive feed that over-samples your weak areas depresses your visible percentage relative to a representative paper — you look worse than you are. Second, once you have learned those re-served items, repeat exposure inflates the percentage — you look better than you are. Either way, a number generated from a non-representative, partly-seen selection is not comparable with a mixed, unseen, blueprint-representative block. The fix is not to distrust the bank; it is to measure readiness somewhere the selection is not optimised around your history.
Blueprint audit: attempts versus official weighting
Do not trust the home-screen average as a coverage claim. Export or note your attempted-question distribution across the content map's areas of clinical practice, and compare it with the official spread rather than with the bank's internal tree. Most candidates find heavy over-attempting in a few comfortable systems and thin coverage of paediatrics, mental health, ethics-and-law, and prescribing/calculation items. An area you have barely attempted is uncovered even if your overall percentage is high. Our blueprint-coverage-matrix method is the tool for this; the GMC content map is the reference to build it against.
Readiness test: the five conditions
A MedRevisions percentage only becomes a readiness signal when all five conditions hold:
- Unseen — items you have not previously attempted in this bank.
- Timed — at the real AKT pace, roughly one minute per item.
- Mixed — blueprint-representative, not a single weak system.
- No assistance — no AI tutor, notes or pausing.
- Adequate sample — a large enough block (a full-length paper's worth) to be stable, not a lucky 20.
If any condition is missing, you have a study metric, not a readiness signal.
Algorithm override rules
Adaptive and "high-yield essential" modes optimise for aggregate efficiency, which means low-volume-but-examinable material can be under-served. Manually force these into your practice regardless of what the feed offers: rarely-surfaced content-map areas; image and data-interpretation items; ethics, law and professionalism; drug calculations and prescribing safety; and any domain the "essentials" mode trims. If the engine will not schedule them, schedule them yourself.
Worked dashboard example: from analytics to quotas
Suppose MedRevisions shows first-attempt accuracy 71%, repeat accuracy 88%, a percentile in the middle band, coverage strong in medicine and surgery but under 30% attempted in paediatrics and psychiatry, and time-per-item running over 80 seconds. Read it as: recognition is inflating the blended score (the 17-point first-versus-repeat gap), two blueprint areas are genuinely uncovered, and pace is unsafe. Next week's quotas follow directly — not a pass prediction:
- 120 first-attempt items split across paediatrics and psychiatry (close the coverage gap).
- 2 timed blocks at strict 60-second pace (fix time-per-item), analysed by error type.
- All missed items re-derived from current UK guidance, not re-read.
- 1 unseen mixed block elsewhere to check the trend without the bank's selection.
Every number here is a task derived from a gap, not a forecast of the result.
Reading your results: three mistakes this audit is designed to stop
First, celebrating a rising blended percentage that is really repeat accuracy — always read first-attempt and repeat separately. Second, treating the vendor's coverage tree as the GMC blueprint — audit attempts against the official map. Third, accepting a predicted score as a verdict — it is a modelled hypothesis, and only an unseen timed block can test it.
A fourth, quieter mistake is reading the percentile as reassurance. A mid-band percentile against a self-selected user pool tells you where you sit among people who happen to use MedRevisions, not whether you clear the GMC standard, which is set by a criterion-referenced process rather than by a rank. If the pool is strong you may pass while sitting mid-table; if it is weak, the reverse holds. Rank data is motivational, not diagnostic — anchor your decisions to your unseen first-attempt trend against the content map and let the percentile sit in the background.
Time-per-item deserves more weight than most candidates give it. At the AKT's roughly one-minute pace, accuracy achieved at ninety seconds an item is a mirage that will not survive the real clock. If your untimed accuracy is strong but your timed accuracy drops, the problem is retrieval speed, not knowledge, and the fix is timed blocks with a hard per-item cap rather than more content. Read accuracy and time together; neither means much alone.
A seven-day plan: MedRevisions for one job, iatroX for the other
For an international graduate balancing content review with UK exam conventions, give each tool one job and do not let them overlap. No proprietary-algorithm claims are made for either.
| Day | MedRevisions (coverage and correction) | iatroX (unseen measurement) |
|---|---|---|
| Mon | 40 first-attempt items in a weak content-map area | — |
| Tue | Re-derive Monday's misses against UK guidance | — |
| Wed | 40 items, second weak area | — |
| Thu | Ethics, calculation and image items (forced) | — |
| Fri | Error analysis; log localisation ledger | — |
| Sat | — | 90-item unseen, timed, mixed block; analyse |
| Sun | Light review of Saturday's error themes | Log the unseen trend point |
MedRevisions does the coverage-and-correction work its analytics support; iatroX supplies the unseen, timed measurement its own selection cannot, so your trend line is uncontaminated by re-served items.
Decision checklist: continue, supplement, switch or stop
- Continue if coverage is still expanding and your unseen trend is rising.
- Supplement (add an unseen measurement bank) if repeat accuracy is high but unseen accuracy is unknown or flat.
- Switch primary bank only if coverage of your weak content-map areas is genuinely thin here — not out of novelty.
- Stop new questions when fresh blocks surface no new gaps and the unseen trend has plateaued safely; move to full timed papers.
Bottom line
MedRevisions is a strong, high-volume AKT and PLAB 1 bank with more analytics than most, and its weakness-focused feed is good teaching. Just remember what its numbers are and are not: first-attempt accuracy and an unseen trend are signal; blended percentages, percentiles and predicted scores are context at best. Use its analytics to fix coverage and errors, and measure readiness on unseen blocks elsewhere.
Frequently asked questions
Is MedRevisions enough for UKMLA on its own? For the applied knowledge test, a bank of this size worked to genuine coverage can be sufficient for many candidates, provided you audit attempts against the GMC content map and measure readiness on unseen timed blocks. It is not sufficient for the UKMLA as a whole, because it does not reproduce the CPSA, which needs structured station practice with a real observer.
Which UKMLA component does MedRevisions not reproduce well? The CPSA. MedRevisions is a knowledge bank for the AKT/PLAB 1 route (vendor-reported), and no MCQ product trains the OSCE-style clinical and professional skills the CPSA assesses. Its written explanations also cannot substitute for observed feedback on history-taking, examination and communication.
How many MedRevisions questions should I complete per day for UKMLA? There is no universal number; set the quota from your gaps, not from a target count. In an active phase, 40–60 first-attempt items concentrated in weak content-map areas, fully analysed, beats a larger volume skimmed. Once coverage is complete, shift from daily new questions to timed full-length papers and error analysis.
When should I stop using MedRevisions and move to mixed mocks? When your first-attempt coverage across the content map is broadly complete and your bank percentage has plateaued, move the centre of gravity to unseen, timed, mixed mocks. Continue to use MedRevisions to correct the errors those mocks reveal, but stop treating a rising bank percentage as progress once it is being driven by repeats.
How should I combine MedRevisions with iatroX without duplicating practice? Give each a single non-overlapping job: MedRevisions for volume, coverage and correction; iatroX for unseen, timed, mixed measurement of transfer. Never re-answer an item you have already seen in one bank inside the other — that duplicates content and corrupts your percentage. Our two-Q-bank rule sets out exactly how to add a second bank without destroying calibration.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026. Question counts, prices and feature names are vendor-reported (MedRevisions, 19 July 2026) and change without notice — confirm them on the product page before relying on them. Disclosure: iatroX operates a competing UKMLA question bank; this audit therefore confines iatroX's role to a job MedRevisions does not claim — unseen, timed, mixed measurement — rather than to volume. Corrections are welcome via the feedback route on iatrox.com.
References: MedRevisions UKMLA/PLAB question bank (medrevisions.com); GMC, Medical Licensing Assessment and MLA content map (gmc-uk.org); iatroX, "Your Q-Bank Percentage Is Not Your Exam Score" (https://www.iatrox.com/blog/qbank-percentage-not-your-exam-score); iatroX, "The Two-Q-Bank Rule" (https://www.iatrox.com/blog/the-two-q-bank-rule-how-to-add-a-second-bank-without-duplicating-questions-or-destroying-calibration); iatroX comparison hub (https://www.iatrox.com/compare).