This workflow is for candidates using QBankMD as their MCCQE Part I bank who want to get real value from its AI assistant without letting it become an answer-leak machine or an unverified oracle. It addresses the single multiple-choice component the exam now has, and it is built around one limitation you must respect: QBankMD's AI is a vendor-reported chatbot, not an examiner and not a guideline database, so every high-stakes output needs a commitment step before it and a verification step after it.
What QBankMD offers for MCCQE Part I right now
Figures below are vendor-reported from the QBankMD site and last checked on 19 July 2026. Confirm them on the product page before you rely on any of them.
| Item | What the site states (vendor-reported, 19 July 2026) |
|---|---|
| Question count | Around 3,900 practice questions (the site also uses "3,000+") |
| AI assistant | An "intelligent chatbot" to ask about exam content, request clarifications and explanations; roughly CA$10 of free AI credits (about 2,000 queries) advertised |
| Adaptive claim | An "adaptive learning algorithm" is listed as included; the mechanism is not specified |
| Access model | Monthly to annual plans in Canadian dollars; longest advertised term is 12 months |
| Indicative pricing | From roughly CA$40/month on the annual plan up to about CA$135 for a single month |
| Free trial | 20 questions |
Be calibrated about two of these. The "AI assistant" is a general chatbot you converse with — useful, but it can be confidently wrong and is not grounded in a verified Canadian guideline set unless the vendor says so. The "adaptive learning algorithm" is a vendor claim with no published description; treat it as marketing until you can see what it reweights. None of this makes QBankMD a poor bank; it means the discipline has to come from you.
The exam you are actually preparing for
Since April 2025 the MCCQE Part I has been entirely multiple-choice — the clinical decision-making cases were removed — delivered by computer as roughly 230 single-best-answer MCQs in two sections of about 115 each, with around two hours and forty minutes per section. Confirm the current count and timing on mcc.ca. The content is anchored to the Medical Council of Canada's Objectives for the Qualifying Examination, spanning dimensions of care and CanMEDS-mapped physician activities across medicine, surgery, paediatrics, obstetrics and gynaecology, psychiatry, and population health, ethics and law. Those objectives and the MCC's practice materials are the authoritative reference; an AI chatbot's answer is not, however fluent it sounds.
A prompt taxonomy for missed questions
Do not ask the assistant "why is this the answer?" — that invites a fluent restatement of the key you have already seen. Ask structured questions that make the model do reasoning you can check. Six prompt families cover almost everything:
- Mechanism — force the underlying physiology or pharmacology.
- Discriminating features — what separates the right answer from the closest wrong one.
- Option elimination — why each distractor fails, one at a time.
- Guideline verification — which Canadian guidance applies, and its date.
- Counterfactual — change one stem detail and ask how the answer changes.
- Retrieval testing — have the model quiz you on the same principle without the original stem.
Twenty-five reusable prompts, grouped by error type
Keep prompts generic so you never paste proprietary question text into a chatbot. Group them by the six error types you will code most often.
Knowledge-gap errors
- "Explain the mechanism behind [condition/finding] in three sentences, then state the single fact I most likely did not know."
- "Give me the one-line rule that governs this decision, with no caveats."
- "List the three highest-yield facts about [topic] for a Canadian licensing exam."
- "What is the classic triad or key feature that should have flagged [diagnosis]?"
Reasoning errors 5. "Compare [right answer] and [closest distractor]: what single discriminating feature decides between them?" 6. "Walk me through the decision tree from presentation to answer, one branch at a time." 7. "What assumption in my reasoning would lead me to pick [distractor], and why is it wrong?" 8. "Rank these options from most to least appropriate and justify the ordering."
Data and interpretation errors 9. "Talk me through how to read this [ECG/lab pattern/imaging description] step by step." 10. "Which single value or finding here is the decision-maker, and why?" 11. "What normal range applies, and how far outside it is clinically meaningful?" 12. "If this lab trend were moving the other way, how would management change?"
Management-sequencing errors 13. "What is the next best step, and what must be done before it?" 14. "Distinguish first-line, second-line and definitive management for [condition]." 15. "What would make this an emergency requiring an immediate action rather than investigation?"
Jurisdiction errors 16. "State whether this answer reflects Canadian practice specifically, and name the Canadian guideline and its year." 17. "How, if at all, does Canadian screening or first-line therapy for [condition] differ from other countries?" 18. "Is this drug or pathway the Canadian standard, or are you generalising?"
Communication, ethics and population-health errors 19. "What is the professionally and ethically correct action here under Canadian norms?" 20. "How should consent, capacity or confidentiality shape this answer?" 21. "What population-health or preventive dimension does this question test?"
Retrieval and consolidation prompts 22. "Quiz me with a new scenario testing the same principle, without repeating the original." 23. "Give me three variations on this concept and let me answer before you confirm." 24. "Summarise what I got wrong in one corrected rule I can memorise." 25. "Two weeks from now, what should you ask me to check I have retained this?"
The anti-answer-leak rule
The assistant is worthless if you open it before committing. Rule: write your chosen answer and a one-line rationale first — on paper or in the misconception log — then open the tutor. Consult the AI while still deciding and you are not practising retrieval; you are outsourcing it, and your first-attempt accuracy becomes meaningless. Commit, then interrogate.
The jurisdiction check prompt
International graduates fail MCCQE items on jurisdiction more than on knowledge, and a fluent chatbot will happily give a British or American answer in Canadian clothing. Keep one reusable prompt to hand: "Does this answer reflect the MCCQE Part I Canadian jurisdiction specifically? Name the Canadian guideline or authority it rests on and the date of that guidance. If you are generalising from non-Canadian sources, say so explicitly." If the model cannot name a Canadian source and date, treat the point as unverified until you check it.
The misconception record
For every genuine miss, log four short lines and nothing more: the incorrect rule you were using, the corrected rule, one transfer question that tests the same principle differently, and a review date. Four lines beats four paragraphs, because you will actually re-read them.
The weekly verification sample
Because the AI can hallucinate citations and sound confident while wrong, audit it. Once a week, take a small random sample of its substantive claims — five is plenty — and check each against the MCC objectives, a current Canadian guideline or a primary source. Log every discrepancy. If the assistant drifts on jurisdiction or invents references, downgrade your trust and lean harder on official material. This is the habit any responsible reader should apply to any AI exam tutor.
Worked example: a seven-day pattern for an international graduate
An international graduate, six weeks out, strong on medicine but shaky on Canadian conventions. QBankMD does one job — questions plus AI-assisted error analysis — and iatroX does one job: unseen transfer practice and a fresh timed baseline, with no proprietary-algorithm claim on either side.
- Monday: 40-item mixed QBankMD block. Commit every answer before opening the assistant. Interrogate misses with mechanism and discriminating-feature prompts.
- Tuesday: run the jurisdiction-check prompt on every J-coded miss from Monday; verify two against a named Canadian guideline yourself.
- Wednesday: transfer day — a fresh iatroX item on each principle before revisiting the QBankMD original; log which survive.
- Thursday: 40-item timed QBankMD block; use option-elimination prompts only, to sharpen distractor logic under pace.
- Friday: weekly verification sample — audit five assistant claims against official sources; update the misconception log.
- Saturday: a longer timed mixed block, then a short unseen iatroX set to confirm gains are not bank-specific.
- Sunday: rest or light review of the misconception list.
Decision checklist: continue, supplement, switch or stop
- Continue if committed-first accuracy is rising and your weekly verification sample shows the assistant is mostly reliable on jurisdiction and citations.
- Supplement with unseen items if your percentage climbs but fresh-item performance stalls, or you cannot tell recognition from recall.
- Switch primary bank only for a measured reason — a domain genuinely under-covered against the MCC objectives, or an assistant that repeatedly fails your verification sample — not because the AI feels impressive.
- Stop new questions when coverage, unseen timed performance, pacing, retention and calibration are all met.
Bottom line
QBankMD pairs a sizeable MCCQE Part I bank with an AI assistant that can genuinely help — if you cage it. Commit your answer before you open it, drive it with structured prompts rather than "why is this right?", force it to declare Canadian jurisdiction and dates, and audit a weekly sample against official sources. Treat the "adaptive" label as an unverified vendor claim. Do that and the AI sharpens your reasoning; skip it and it just leaks answers in fluent prose.
Frequently asked questions
Is QBankMD enough for MCCQE Part I on its own? As a primary bank with an AI layer it can carry most of the load for a well-prepared candidate, provided you calibrate against the MCC's own materials and verify the assistant rather than trusting it. But the AI does not make one bank sufficient: a chatbot you have talked through every question with stops being an unseen readiness signal. Add official practice material and a fresh, unseen source for honest measurement.
Which MCCQE Part I component does QBankMD not reproduce well? The exam is now wholly multiple-choice, so nothing structural is missing; the weak spot is grounding. QBankMD's assistant is a general chatbot (vendor-reported), not a verified Canadian-guideline engine, so it does not reliably reproduce the jurisdictional precision the MCC objectives demand — which is why the jurisdiction-check prompt and weekly verification sample exist.
How should I verify QBankMD AI answers for MCCQE Part I? Commit your own answer first, then ask the assistant to name the Canadian guideline behind its response and the date of that guidance; each week, check a small random sample of its substantive claims against the MCC objectives or a current primary source and log discrepancies. If it cannot cite a Canadian source, treat the claim as unverified until you confirm it.
When should I stop using QBankMD and move to mixed mocks? Move to mixed, timed blocks once each domain clears its floor and your committed-first accuracy is stable, rather than waiting to finish the bank; the assistant is a learning tool for the build phase, not something you should be consulting during exam-simulation blocks. Sit those mocks on items you have not talked through with the AI.
How should I combine QBankMD with iatroX without duplicating practice? Assign each a single job: QBankMD for questions and AI-assisted error analysis, iatroX for unseen transfer practice and a fresh timed baseline, so you never re-answer an item whose answer the chatbot has already handed you. This preserves an untainted measurement layer — the point of the two-bank rule — without either platform claiming a proprietary algorithm.
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, AI credits, the "adaptive" claim) are vendor-reported from the QBankMD site on that date, are unverified by iatroX, and change without notice — confirm the current figures on the product page. Disclosure: iatroX operates a competing MCCQE Part I bank and a Socratic Tutor, so this article confines iatroX's role to unseen transfer practice and timed measurement — jobs QBankMD's own workflow does not claim — and makes no proprietary-algorithm claims about either platform. Corrections are welcome via the feedback route on iatrox.com.
References: Medical Council of Canada, MCCQE Part I and multiple-choice examination pages and the Objectives for the Qualifying Examination (mcc.ca); QBankMD product and blog pages (qbank.md). Related iatroX reading: the MCCQE Part I bank landing page (iatrox.com/mccqe1), "Your Q-Bank Percentage Is Not Your Exam Score", the guide to auditing an AI medical-exam tutor (grounding, answer leakage, hallucinations and retention), and the two-Q-bank rule.