This workflow is for GP trainees and IMG GPs who already use, or are about to use, AKTPrep's AI tutor and want a repeatable routine — not another product ranking. It turns each missed MRCGP AKT question into a short, disciplined interrogation that pins down the discriminator, forces a jurisdiction and currency check, and produces one durable correction. The principal limitation it manages: an AI tutor's fluent prose is a draft, most fragile on the statistics and organisational strands, so the workflow builds verification in rather than bolting it on.
What AKTPrep offers for the MRCGP AKT right now
These vendor-reported details were last checked on 19 July 2026; confirm the live figures on the AKTPrep product page before relying on them.
| Feature | Vendor-reported detail (verify on product page) |
|---|---|
| Question volume | 2,000+ SBA and EMQ items across 28 areas |
| AI tutor | Follow-up questions and feedback on wrong answers; NICE/UK-guideline-focused |
| Adaptive engine | "Weakness engine" ranking topics by mastery, with weak-topic retries |
| Mocks | Two full timed mocks with score prediction |
| Statistics | Research & Statistics module: 40 lessons, calculators, timed mock |
| Pricing | 1 month £14.95; 3 months £35.97; 6 months £49.99 (vendor-reported) |
The tutor and weakness engine are the point of the product, so the return on your subscription depends on how well you prompt and verify, not on how many items you grind.
The MRCGP AKT blueprint you are actually preparing for
From October 2025 the AKT is 160 single-best-answer items in 2 hours 40 minutes, four sittings a year at Pearson VUE, at about one minute per item. The blueprint is roughly 80% clinical, 10% evidence-based practice (statistics and critical appraisal) and 10% organisational (UK GP administration and regulation), all in a UK GP context and all currency-dependent. The RCGP AKT content guide and official sample questions are the reference standard; treat the tutor's output as a hypothesis to be checked against them, never as the authority itself.
A prompt taxonomy for missed questions
Every good prompt does one job. Six job types cover almost every AKT miss: mechanism (why is this the answer at the level of physiology or pharmacology?), discriminating features (what single feature separates the best answer from the runner-up?), option elimination (why is each distractor wrong, one at a time?), guideline verification (which current UK guideline says this, and when was it updated?), counterfactuals (change one detail — age, renal function, a red flag — and does the answer change?), and retrieval testing (ask the tutor to quiz you on the same principle in a new scenario). Keep prompts reusable and generic; never paste proprietary question text into a note you keep or share.
Twenty-five reusable prompts, grouped by error type
Diagnose why you missed the item first, then use the matching prompt set. The six error types below map onto the exam's strands.
1. Knowledge gap (you did not know the fact or mechanism):
- "Explain the mechanism behind the correct answer in two sentences, then name the source."
- "What is the one fact I needed to know here, stated as a rule?"
- "Give me a memory hook for this rule that does not distort it."
- "Quiz me on this mechanism with a different patient in one week."
2. Discriminator miss (you were down to two options):
- "What is the single discriminating feature between the correct answer and my choice?"
- "List the features that would have pointed to my answer instead — when is it right?"
- "Rank all five options from most to least likely and justify the order."
3. Stem misread or data misinterpretation:
- "Which specific words in the stem determine the answer?"
- "Re-read the data: what did I over-weight or ignore?"
- "Rewrite this stem so my answer becomes correct — what changed?"
4. Guideline-currency error (you knew older guidance):
- "Which current UK guideline governs this, and what is its publication or review date?"
- "Has first-line management here changed in the last three years? State old versus current."
- "What is a plausible out-of-date answer an examiner would use as a distractor?"
- "Give me the current CKS position in one line so I can verify it."
5. Statistics or critical-appraisal error:
- "Show the calculation step by step and state the formula."
- "Convert this 2×2 table into sensitivity, specificity, and both predictive values."
- "Explain the difference between relative and absolute risk using these numbers."
- "Interpret this confidence interval and p-value in one plain sentence each."
- "What study design would answer this question, and what is its main bias?"
6. Jurisdiction or organisational error (UK GP admin and regulation):
- "Does this answer reflect UK general practice specifically? Where would it differ elsewhere?"
- "State the current UK rule on this administrative or regulatory point and cite the body."
- "Walk through the correct referral or certification pathway step by step."
- "Which regulator or guidance owns this rule — RCGP, GMC, CQC, DVLA or DWP?"
Cross-cutting retrieval and calibration:
- "Set me one transfer question on this principle without repeating the original."
- "How confident should I be in your last answer, and what could make it wrong?"
The anti-answer-leak rule
The workflow only works if you commit first. Before you open the tutor on any item, write down your answer and a one-line rationale. If the tutor reveals or implies the answer before you have committed, you lose the single most valuable piece of diagnostic information — what you actually believed. Treat "answer first, tutor second" as non-negotiable, exactly as you would in the Socratic Tutor in iatroX, where the interrogation is meant to precede the reveal.
The reusable jurisdiction-and-currency check
Bind one prompt to every management, prescribing or organisational item: "State whether this answer reflects the MRCGP AKT UK general-practice jurisdiction, name the specific guideline, and give the date of that guidance." Then open the named source and confirm it. A model cannot know what changed after its training data froze, so this check is where you catch the stale-but-plausible answer that the examiners deliberately plant. Calibrating automated feedback before you trust it is a discipline in its own right, covered in the iatroX pillar on how to calibrate AI-graded feedback.
Create a misconception record
For every genuine miss, write four lines and no more: the incorrect rule you were carrying, the corrected rule in one sentence, one transfer question that tests the same principle in a new scenario, and a review date. Four lines force you to distil, and distillation is what transfers; transcribing the tutor's whole paragraph does not. Over a few weeks this record becomes your highest-yield revision document — a personal list of the exact rules that cost you marks, each already paired with a re-test.
A weekly verification sample
Once a week, take a small random set of the tutor's outputs from that week — five is enough — and independently check each against a primary source: CKS, NICE, SIGN, a SmPC, or the relevant regulator. Log any discrepancy. Two purposes: you correct real errors before they calcify, and you build a running sense of how far to trust this tutor on each strand. If your log shows the tutor is reliable on clinical items but wobbly on statistics, that is not a reason to abandon it — it is a reason to lead with the statistics module and use the tutor only to check your working.
A worked seven-day plan
One week, AKTPrep for interrogation and remediation, iatroX for unseen timed measurement, no proprietary-algorithm claims.
- Day 1: Sit an AKTPrep timed mock cold for a clean baseline.
- Day 2: Take your ten worst misses, classify each by error type, and run the matching prompts under the answer-first rule. Write a four-line misconception record for each.
- Day 3: Statistics module plus the statistics prompt set until NNT, likelihood ratios and predictive values are automatic.
- Day 4: Organisational items with the jurisdiction-and-currency check on every one; verify each rule against the live UK source.
- Day 5: Run a fresh, timed mixed block in iatroX; the unseen items measure transfer, not recall.
- Day 6: Open every iatroX miss, commit first, verify, and re-test your two weakest domains; run your weekly verification sample.
- Day 7: Second timed mock; compare it with your unseen iatroX percentage, remembering that a bank percentage is not an exam score.
Continue, supplement, switch or stop
Continue if your misconception records are shrinking your repeat errors and your unseen scores track your mocks. Supplement if clinical is solid but statistics or organisational stays weak. Switch the tool if your weekly verification log fills with citations that do not check out. Stop adding tools once you are meeting domain floors on unseen timed blocks — more banks then add duplication, not signal.
The bottom line
AKTPrep's tutor rewards discipline. Prompt it by error type, force a jurisdiction and currency check on anything that can go stale, distil every miss into four lines, and audit a weekly sample against primary sources. Do that and the tutor becomes a genuine coach for the clinical majority of the AKT while you keep firm control of the statistics and organisational strands. Skip the verification and its confident prose will quietly cost you the marks that separate a pass from a resit.
Frequently asked questions
Is AKTPrep enough for MRCGP AKT on its own? Its vendor-reported 2,000+ items, tutor and two mocks can cover the clinical majority, but "enough on its own" depends on the statistics and organisational strands, which need primary-source verification and unseen timed practice. For most candidates AKTPrep is a strong core rather than a complete solution.
Which MRCGP AKT component does AKTPrep not reproduce well? The two non-clinical strands carry the most risk, because AI explanations of statistics and UK-specific organisational rules are the hardest to trust unverified, and no single practice tool fully reproduces the timed, 160-item exam experience — which you approximate only with unseen timed blocks.
How should I verify AKTPrep AI answers for MRCGP AKT? Commit to your answer first, prompt for the discriminator and the named source, then verify anything currency-dependent, statistical or organisational against the live primary source and record the verified rule. Run a five-item verification sample weekly and log discrepancies so you know how far to trust each strand.
When should I stop using AKTPrep and move to mixed mocks? Once you meet a floor in every blueprint domain on unseen questions and your pacing sits near one minute per item, shift to predominantly mixed, timed blocks. Topic drilling with a tutor builds foundations; the exam tests transfer under time.
How should I combine AKTPrep with iatroX without duplicating practice? Give each tool one job: AKTPrep for interrogation and remediation, the free UK-core iatroX bank for unseen timed measurement. Never re-test an item you have already dissected. See the comparison hub for how the tools differ in role.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026. AKTPrep figures are vendor-reported and were taken from the AKTPrep product pages on that date; confirm current counts, prices and features on the product page. Disclosure: iatroX operates a competing MRCGP AKT question bank and Socratic Tutor; this workflow confines iatroX's role to unseen, timed measurement and independent verification — the check AKTPrep does not claim to be for its own output. Corrections are welcome via the feedback route on iatrox.com. References: RCGP AKT content guide and official sample questions (rcgp.org.uk); NICE, CKS and SIGN; the electronic Medicines Compendium (eMC) SmPCs; AKTPrep product pages (aktprep.uk); and the iatroX MRCGP AKT bank.