This audit is for standard-pathway candidates using — or considering — AusProEd (ausproed.com) for the AMC MCQ, and specifically for its analytics and its vendor-described adaptive feed. In short: AusProEd offers a large, Australian-authored bank with performance analytics and a stated ability to "adapt to your performance". That is useful for finding weak spots. Its principal limitation is definitional: an adaptive feed is optimised to keep you at the edge of your ability, which makes your raw percentage on adaptively selected items a poor, non-comparable estimate of exam readiness.
Current state (verified 19 July 2026)
Vendor-reported from the product's own pages on the date shown; confirm on the site.
| Item | What the product page shows |
|---|---|
| Exam covered | AMC MCQ (AMC Computer Adaptive Test) |
| Question volume | 3,000+ practice questions, written and reviewed by Australian doctors (vendor-reported) |
| Adaptive feature | Vendor states the platform "adapts to your performance, focusing on areas where you need improvement" — mechanism not detailed; treat as vendor-reported, not a verified algorithm |
| Analytics | Performance analytics to identify strengths, weaknesses and readiness (vendor-reported) |
| Explanations | Detailed explanations that teach underlying concepts (vendor-reported) |
| Content updates | Refreshed quarterly to reflect current Australian guidelines (vendor-reported) |
| Community | Forum access (vendor-reported) |
| Access and price | Plans from 1 month (about A$125) to 6 months (about A$450, listed down from A$650); 2- and 3-month plans between (vendor-reported) |
One caution up front: AusProEd describes an adaptive capability but does not publish how it selects items. Do not assume a validated psychometric engine; treat "adaptive" as a vendor-reported feed that steers you toward weak areas, and read the analytics accordingly.
Exam anchor
The AMC MCQ is 150 single-best-answer questions (one of five) in one 3.5-hour computer-administered session at Pearson VUE, blueprint-weighted about 30% Adult Health Medicine, 20% Surgery, 25% Women's Health, 12.5% Child Health, 12.5% Mental Health, with Population Health and Ethics distributed, and it is the gateway to the AMC Clinical Examination. The AMC calls the exam itself a "Computer Adaptive Test", but its public page does not confirm a difficulty-adapting algorithm; plan around a fixed-length computer-administered MCQ and verify the mechanism on the AMC specifications. Note the distinction that matters for this article: whatever the exam's mechanism, a bank's "adaptive" feed is a separate, vendor-designed thing, and the two should not be conflated.
Define every metric before you trust it
Analytics dashboards show numbers that look comparable but are not. Define each before you act on it.
- First-attempt accuracy — your percentage on items the first time you see them. This is the only accuracy figure that approximates learning; guard it.
- Repeat accuracy — your percentage on items you have already attempted. It rises mechanically with exposure and tells you little about readiness.
- Percentile / peer rank — your standing against other users of this platform, a self-selected group, not against the AMC standard.
- Predicted score — any single "you will pass" number is a modelled estimate on the vendor's own content; treat it as motivational, not diagnostic, and never plan around it.
- Coverage — how much of the bank you have attempted, which is completion, not blueprint coverage.
- Difficulty — the platform's own label, not an official calibration.
- Time per item — arguably the most useful metric here, because pace transfers directly to the real 84-seconds-an-item demand.
Selection bias: why an adaptive feed distorts your percentage
Here is the core of what the engine is "optimising". If the feed preferentially serves items in your weak areas — which is what "focusing on areas where you need improvement" means — then the pool you are answering is deliberately skewed hard and skewed toward your gaps. Your accuracy on that pool will read lower than your true standing across a balanced blueprint, and it is not comparable week to week, because the mix keeps changing beneath you. The opposite failure occurs if you let the feed drift toward mastered material: your percentage climbs while your real weak areas go unpractised. Either way, an adaptively selected percentage is the wrong denominator for a readiness judgement. The only percentage that supports comparison is one from a fixed, balanced, unseen block — the reason we keep returning to unseen measurement.
Blueprint audit: check your distribution, not your average
Do not trust the home-screen average. Export or tally your attempted-question distribution and compare it with the AMC weighting: are about 30% of your items Medicine, 25% Women's Health, and so on, or has the adaptive feed pulled you into a narrow band of your weakest topics while a quarter-weighted strand goes thin? An adaptive feed optimises for your improvement curve, not for blueprint balance, so the two diverge unless you force the correction. Our guide on building a blueprint-coverage matrix gives the template.
Readiness test: the five conditions for a credible signal
A number only estimates readiness if all five conditions hold at once: the items are unseen (not previously attempted on this platform), the block is timed at exam pace, the mix is mixed across the full blueprint rather than adaptively narrowed, you take it with no assistance (no explanations mid-block, no pausing to look things up), and the sample is large enough — a few dozen items at minimum — that one lucky run does not move it. AusProEd's adaptive feed, by design, breaks the "mixed" and often the "unseen" conditions, so its live percentage is not that signal. Construct the signal deliberately, on a fixed unseen block, and treat everything the dashboard shows as diagnostic colour around it.
Algorithm override rules
Because the feed will under-serve some material, override it on a schedule. Force low-volume, high-consequence content the engine may not surface often enough: image and data interpretation, ethics and Population Health, calculations and prescribing safety, Mental Health law, and any strand your distribution audit shows below its blueprint weight. Once a week, ignore the "recommended" queue entirely and build a manual, balanced, timed set across all six strands. The override is not fighting the tool; it is supplying the blueprint balance an improvement-seeking feed is not designed to guarantee.
Worked dashboard example
Suppose the dashboard reads: overall accuracy 71%, first-attempt 62%, repeat 88%, "predicted: likely pass", coverage 74%, and time-per-item 96 seconds. Read it like this. The 71% is inflated by the 88% repeat figure — lean on the 62% first-attempt number. Coverage 74% is completion, not blueprint coverage, so check the distribution. Time-per-item at 96 seconds is over the ~84-second budget, so pace is a live risk. Ignore "likely pass" entirely. Next week's quotas write themselves: a fixed, unseen, timed 40-item block across all strands to get a comparable accuracy; a forced set in your two below-weight strands; and every block run at an 84-second ceiling to pull your pace down. No pass prediction, just next actions.
Worked example: a seven-day plan
An overseas-trained candidate balancing content review with Australian conventions, using AusProEd for one job — targeted weak-area practice with its analytics — and iatroX for a separate job: unseen, timed measurement. No proprietary algorithm is assumed on either side.
- Day 1: Let the AusProEd feed run for 40 items; note which strands it served and which it ignored.
- Day 2: Override the feed — a manual, balanced 40-item set across all six strands, timed; log misses.
- Day 3: Review misses; check each localisation-sensitive answer against Therapeutic Guidelines or the Australian Medicines Handbook.
- Day 4: A 40-item unseen, timed iatroX block, mixed, no assistance — your comparable readiness number.
- Day 5: Force the low-volume material — images, ethics, calculations, Mental Health law — the feed under-serves.
- Day 6: 50 items at an 84-second ceiling to train pace; record time-per-item.
- Day 7: Compare this week's unseen accuracy and pace with last week's; reset quotas from the gap, not from the dashboard average.
Decision checklist: continue, supplement, switch or stop
Continue if the analytics are genuinely changing your study — steering review to real weak spots — and your unseen accuracy is rising. Supplement if you have no fixed unseen measure (you need one the adaptive feed cannot provide) or if a distribution audit shows a strand starved; add balanced, unseen blocks. Switch if the "adaptive" feed is opaque, cannot be overridden and is trapping you in a narrow band, or if a jurisdiction sample shows content that has drifted from current Australian guidance despite the quarterly-update claim. Stop adding volume when first-attempt accuracy has plateaued across a balanced blueprint and pace is under control. Judge on measurable gaps, not on the novelty of the dashboard.
Sanity-checking the "updated quarterly" claim
AusProEd states its content is refreshed quarterly to track current Australian guidelines, which is reassuring if true and straightforward to test. Take five items in fast-moving areas — empirical antibiotics, anticoagulation choice, contraception, immunisation catch-up and mental-health law — and check each answer and explanation against Therapeutic Guidelines, the Australian Medicines Handbook and the National Immunisation Program on the day you review them, recording the date you did it. If the bank's answers track the current Australian standard, the update claim is holding for the content you actually rely on; if you find a superseded first-line therapy or an out-of-date schedule, the cadence is not reaching those items, and you should weight your own guideline check above the platform's explanation. Recency is a claim to verify, not a feature to assume, however often a vendor says it refreshes.
Bottom line
AusProEd's engine is optimised to keep serving you your weak areas, which is a good learning behaviour and a bad measurement one. Use its analytics to find gaps and its volume to close them, but do not read an adaptively selected percentage as a readiness estimate. Build that estimate yourself on fixed, unseen, timed, balanced blocks, and treat "predicted pass" as noise.
Frequently asked questions
Is AusProEd enough for AMC MCQ on its own? As a large, Australian-authored bank with analytics it can carry the learning and volume load on its own for many candidates. It is not enough on its own for a trustworthy readiness signal, because its adaptive feed deliberately narrows and re-serves items, which breaks the fixed, mixed, unseen conditions a credible signal needs. Pair it with a separate unseen measure, and add a clinician check for the judgement-heavy ethics content.
Which AMC MCQ component does AusProEd not reproduce well? It does not reproduce a balanced, blueprint-representative test experience, precisely because an improvement-seeking feed skews toward your weak areas; nor does its live percentage reproduce the comparability of a fixed unseen block. It also does not train the separate AMC Clinical Examination. Its analytics are diagnostic, not a substitute for simulation.
How many AusProEd questions should I complete per day for AMC MCQ? Enough to review well — commonly 40 to 60 — but with a weekly override where you ignore the recommended queue and do a balanced, timed set across all strands. Throughput matters less than making sure the feed has not quietly narrowed your practice; the count on the dashboard is completion, not coverage.
When should I stop using AusProEd and move to mixed mocks? Shift the emphasis to fixed, mixed, timed mocks once your first-attempt accuracy has plateaued across a balanced blueprint and your pace sits near 84 seconds an item. Keep AusProEd for targeted weak-area drilling, but move your measurement onto unseen, blueprint-balanced blocks, because that is the only percentage you can compare across weeks.
How should I combine AusProEd with iatroX without duplicating practice? Give each a distinct job: AusProEd for adaptive weak-area practice and diagnosis, iatroX for fixed, unseen, timed measurement of transfer. Do not feed the same topics through both on the same day and count them twice, and never let an adaptive learning feed double as your grader. The two-Q-bank rule sets out how to keep the second source additive and your calibration intact.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; question counts, prices, the adaptive claim, update cadence and features are vendor-reported from the product's own pages on that date and change without notice — verify each on ausproed.com. AusProEd describes an adaptive capability but does not publish its selection mechanism; we treat "adaptive" as vendor-reported, not a verified algorithm, and iatroX makes no proprietary-algorithm claim of its own. iatroX operates a competing AMC MCQ bank, which we disclose; this audit confines iatroX to the job the audited product does not claim — independent, unseen measurement. Corrections are welcome via the feedback route on iatrox.com.
References: Australian Medical Council — AMC CAT MCQ examination format and MCQ specifications (amc.org.au); AusProEd AMC MCQ QBank product and pricing pages (ausproed.com, vendor-reported); iatroX AMC MCQ bank (iatrox.com/australia/exam/au-amc); iatroX comparison hub; "Your Q-Bank Percentage Is Not Your Exam Score"; blueprint-coverage matrix; two-Q-bank rule.