This workflow is for CCFP candidates using SampChamp who want each missed case to change their knowledge, not just their score. SampChamp is an AI-marked SAMP practice platform: it grades your free-text answer fields against a rubric and returns detailed feedback in minutes, which is genuinely useful for the written component. Its principal limitation is the one every AI feedback tool shares — the marking and explanations need verifying against a primary source — and, structurally, its AI patient practice is not the live CFPC Simulated Office Oral. Here is how to use it well.
Current-state box (checked 19 July 2026)
Figures are vendor-reported by SampChamp and were last checked on 19 July 2026.
| Item | What SampChamp reports (vendor-reported) |
|---|---|
| What it is | AI-marked CCFP SAMP practice with timed practice exams |
| Cases | 300+ evidence-based cases, described as covering all priority topics ("105") |
| AI feedback | Automated marking with "detailed, personalized feedback for every answer field," delivered in under ten minutes |
| SOO feature | "SOOPER" AI-patient practice (12 AI patients), included at the Pro tier |
| Extras | Timed 20- and 40-case exams, analytics, progress tracking, Anki export |
| Price | Basic US$297, Pro US$447 (one-time); free trial of 5 cases |
| Access | Season access through to the exam date (stated as 14 October 2026) |
| CCFP support | SAMP-focused, with AI-patient SOO practice at Pro |
Exam anchor: what you are preparing for
The CCFP examination has a written component (SAMPs, four hours, transitioning to multiple-choice and short-menu formats across 2026–2027 — up to 25% of cases in 2026, all cases by 2027; verify the split on cfpc.ca) and five 15-minute Simulated Office Orals conducted virtually. SampChamp's core product trains the written SAMP reasoning; its "SOOPER" feature offers AI-patient rehearsal, which is a practice aid, not the live examiner-marked SOO. Keep that boundary clear as you build the workflow.
The prompt taxonomy
An AI tutor is only as good as what you ask it. Structure every review around six prompt types, each targeting a different failure mode:
- Mechanism — why the right answer is right at the level of pathophysiology or evidence.
- Discriminating features — what separates the correct diagnosis or plan from its nearest rival.
- Option elimination — why each wrong answer field is wrong, one at a time.
- Guideline verification — which Canadian guideline supports the answer, and its date.
- Counterfactuals — what single change to the case would change the management.
- Retrieval testing — a fresh transfer question that forces you to apply the corrected rule.
Twenty-five reusable prompts, grouped by error type
Keep these generic — never paste proprietary question text into a tool. Each is reusable across cases.
Mechanism gaps
- "Explain the underlying mechanism that makes the correct management the priority here, in three sentences."
- "What is the physiological reason the first-listed option is wrong despite looking plausible?"
- "Which single pathophysiological fact, if I had known it, would have led me to the right answer?"
- "Summarise the mechanism as a one-line rule I could apply to a different presentation."
Discriminating features
- "List the two features that distinguish this diagnosis from its closest differential in primary care."
- "What finding in the stem should have moved me from my answer to the correct one?"
- "If the patient were five years older, which discriminating feature would matter most?"
- "Give me the single most specific feature that rules this condition in."
Option elimination
- "For each answer field I completed, state whether it was essential, acceptable or wrong, and why."
- "Which of my listed actions was unsafe or contraindicated, and on what basis?"
- "Rank my management steps in the correct order and explain any I misordered."
- "Which option is a common distractor that scores zero, and why is it tempting?"
Guideline and jurisdiction verification
- "Which current Canadian guideline supports this answer, and what is its publication date?"
- "Does this recommendation differ between Canadian and US practice? If so, how?"
- "State the correct Canadian screening interval or target for this condition and cite the source."
- "Flag anything in this explanation that may be out of date and tell me what to verify."
Counterfactuals
- "What one change to this case would make my original answer correct?"
- "If the presenting complaint were the same but the patient were pregnant, what changes?"
- "At what threshold would the management escalate, and why?"
- "Which comorbidity, if added, would contraindicate the recommended step?"
Retrieval testing and transfer
- "Write me a new, different case that tests the same rule I just got wrong."
- "Give me a one-line stem where the corrected rule is the key, and ask me for the answer."
- "Quiz me on this concept again in a way that does not reuse this vignette."
- "What is the single sentence I should remember from this case?"
- "Turn my corrected rule into a flashcard prompt and expected answer."
The anti-answer-leak rule
The discipline that makes any of this work: commit before you consult. Write your full answer and a one-line rationale for every field before you open the AI feedback. If you read the marking first, you contaminate the test — you will feel you "knew it" when you merely recognised it. Answer, commit, then verify. This is the same principle our pillar on auditing an AI exam tutor sets out for grounding and answer leakage.
The jurisdiction check
Make one prompt non-negotiable on every guidance-sensitive case: "State whether this answer reflects current CCFP/Canadian practice and give the date and source of the supporting guidance." AI tools trained largely on US material drift to US defaults on screening intervals, thresholds and drug availability. Forcing the tool to declare jurisdiction and date turns a silent error into a visible one you can check.
Create a misconception record
For every genuine miss, log four lines and nothing more:
- Incorrect rule — the wrong thing you believed, in one sentence.
- Corrected rule — the right rule, in one sentence.
- Transfer question — a fresh prompt that tests the corrected rule.
- Review date — when you will re-test it (a fortnight is a good default).
A page of these is worth more than a thousand re-read explanations, because it captures the specific wrong belief and schedules its correction.
The weekly verification sample
Because you are trusting an AI to mark you, audit it. Once a week, take a small random sample — five of its outputs — and check them independently against current Canadian guidance or a primary source. Log any discrepancy: a wrong threshold, an outdated recommendation, a US-jurisdiction slip. If your discrepancy rate is low, trust the tool more; if it is climbing on a topic, stop trusting it there and verify every answer in that domain. Calibrating the marker before you trust the mark is the method in our calibrating AI-graded feedback pillar.
Worked example: a seven-day plan around clinical work
Give SampChamp one job — high-volume AI-marked SAMP feedback on a defined topic — and give an unseen bank the separate job of measuring transfer.
- Monday (1h): Ten SampChamp cases in your weakest cluster; commit answers before opening feedback.
- Tuesday (1h): Run the six-prompt review on Monday's misses; write a misconception record for each.
- Wednesday (45m): Jurisdiction-check every guidance-sensitive answer from Monday; correct any US drift.
- Thursday (1h): Ten fresh SampChamp cases, same cluster, timed, to confirm the corrected rules hold.
- Friday (1h): Switch tools. Unseen, mixed, timed iatroX block across the CCFP blueprint — the cluster now interleaved — reviewed with the Socratic tutor that names the misconception behind each miss.
- Saturday (45m): Weekly verification sample — check five AI outputs against Canadian guidance; log discrepancies.
- Sunday (1h): Live SOO practice with a study partner — the component neither tool marks for real.
No case is tested twice, and no proprietary algorithm is assumed: SampChamp supplies AI-marked written feedback, iatroX supplies unseen measurement and misconception-level review, a human supplies the oral rehearsal.
Decision checklist: continue, supplement, switch or stop
- Continue SampChamp while its AI feedback is teaching you and your verification sample shows a low discrepancy rate.
- Supplement it with unseen mixed measurement and human SOO practice — the two jobs it does not do.
- Switch or restrict it on any topic where your weekly audit shows the AI marking is unreliable.
- Stop adding new cases when unseen mixed blocks are stable and your remaining gaps are SOO communication skills.
Frequently asked questions
Is SampChamp enough for CCFP on its own? No. It is a strong option for high-volume, AI-marked SAMP practice on the written component, and its rapid feedback is a genuine efficiency, but it cannot be enough on its own: the CCFP includes five live, examiner-marked Simulated Office Orals that its AI-patient feature rehearses but does not replicate, and any AI marking needs independent verification against Canadian guidance. Use it as the written-practice engine, then add unseen measurement and human oral practice around it.
Which CCFP component does SampChamp not reproduce well? The live SOO. Its "SOOPER" AI-patient feature is a useful rehearsal aid, but the real SOO is a virtual consultation marked by a trained examiner on interviewing skill across defined phases, and no AI patient reproduces that judgement or scoring. Its strength is the written SAMP and, increasingly, the multiple-choice and short-menu formats; the oral communication component sits outside what it can faithfully assess and needs human-marked practice.
How should I verify SampChamp AI answers for CCFP? Commit your own answer and rationale before opening the feedback, then run a guideline-verification prompt that forces the tool to name the supporting Canadian guideline and its date. Weekly, take a random sample of five AI outputs and check them independently against current Canadian guidance or a primary source, logging any discrepancy. If the discrepancy rate climbs on a topic, stop trusting the tool there and verify every answer in that domain. Verifying the marker before you trust the mark is the whole discipline; an unverified AI score is not evidence of readiness.
When should I stop using SampChamp and move to mixed mocks? Move to mixed, unseen, timed mocks once your single-topic SampChamp cases are consistently strong and your errors are transfer, timing or communication issues rather than knowledge gaps. More single-topic AI-marked practice past that point cannot change your position, because the limiting factor becomes retrieving knowledge in a mixed paper and performing in the oral — neither of which topic-sorted written cases test. Keep SampChamp for targeted re-teaching of any cluster a mock exposes.
How should I combine SampChamp with iatroX without duplicating practice? Give them separate jobs. Use SampChamp for AI-marked written SAMP practice and rapid feedback on a named cluster; use iatroX for the different job of unseen, mixed, timed measurement mapped to the CFPC blueprint, reviewed with a Socratic tutor that names the misconception behind each miss. Never re-test yourself on a case you have already seen in either tool, and read our CCFP content-gap checklist to decide which cluster to work next. The two-Q-bank rule keeps the two from contaminating each other's calibration.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; SampChamp's case counts, AI features, prices and season-access dates are vendor-reported — confirm the current details on sampchamp.ca and the current exam format on cfpc.ca before you rely on them. Disclosure: iatroX operates a competing Canadian question bank with a Socratic tutor mapped to the CFPC objectives; this article confines iatroX's role to unseen blueprint measurement and misconception-level feedback, and neither iatroX nor SampChamp's AI-patient feature is a substitute for the live examiner-marked Simulated Office Oral. Corrections are welcome via the feedback route on iatrox.com.
References: College of Family Physicians of Canada — Certification Examination in Family Medicine, SAMPs and SOOs (cfpc.ca); SampChamp product and pricing (sampchamp.ca); iatroX CCFP exam page (iatrox.com/canada/exam/ca-ccfp); iatroX — "Your Q-Bank Percentage Is Not Your Exam Score"; iatroX framework pillars on auditing an AI tutor and calibrating AI-graded feedback.