Personalised revision should change what you practise next, when you revisit it or how an explanation responds to your reasoning. It should not merely decorate a standard question list with an AI label. For MRCP Part 1, the useful comparison between Medibuddy and Pastest is therefore the learning decision each function supports, not whether either company uses artificial intelligence.
This article is published by iatroX and includes iatroX in the comparison. Vendor descriptions were reviewed on 24 September 2026; the proposed learner exercises below are inspection methods, not results from authenticated testing of the platforms.
Personalisation is several different functions
Automated selection chooses the next question. A topic recommendation tells you where to concentrate without necessarily choosing the question. Spaced review schedules another encounter with something previously studied. Difficulty adjustment changes the challenge level. Conversational tutoring responds to an explanation, misconception or follow-up question. These functions can coexist, but one does not establish the others.
A useful question for any product is: what information enters the decision? Answer accuracy alone cannot distinguish confident understanding from a lucky guess. Time taken may identify hesitation, but it can also reflect interruption. A topic score based on a narrow sample should not be interpreted as complete knowledge of that topic.
The Medibuddy MRCP Part 1 page, checked on 24 September 2026, describes adaptive selection, topic mastery, spaced repetition and the option to switch to a traditional filtered bank. Pastest's Part 1 page, reviewed on the same date, describes analytics, tutor mode and Ask Pastest alongside its questions and textbook. It would therefore be inaccurate to dismiss Pastest as a static bank or imply that AI support belongs only to iatroX.
A feature-definition matrix
This matrix distinguishes published functions from questions that require an actual demonstration. It is not a quality ranking or a claim that unpublished features are absent.
| Function | Medibuddy's public description, 24 September 2026 | Pastest's public description, 24 September 2026 | What to inspect in practice |
|---|---|---|---|
| Next-question selection | Adaptive targeting of weaknesses | Question practice and analytics; exact selection logic not established here | Can you see why a topic appears? |
| Scheduled return | Spaced repetition explicitly described | Exact review scheduling not established from this product page | Does a missed concept return later? |
| Learner control | Adaptive and traditional modes; filters described | Inspect available question-selection controls | Can you preserve mixed, unseen practice? |
| Explanation support | Short and extended explanations described | Textbook, tutor mode and Ask Pastest described | Does the explanation address your actual mistake? |
| Confidence-sensitive review | Exact confidence-entry behaviour not established | Exact confidence-entry behaviour not established | Can a correct guess enter the review queue? |
| Predictive interpretation | Mastery estimates are product outputs | Analytics are product outputs | Is any exam-outcome prediction independently validated? |
Per iatroX product information, September 2026, its question banks use adaptive sequencing and spaced repetition, while Socratic Tutor starts from the attempted question and asks follow-ups intended to identify the misconception. Those are design features, not proof that every recommendation or explanation is correct. The same inspection questions apply to iatroX's Tutor.
Follow one learner through a week
Consider a fictional trainee, Priya, whose cardiology answers are often correct but slow, whose pharmacology errors cluster around mechanisms, and whose basic-science performance varies substantially by topic. Her problem is not simply that all three dashboard bars should increase.
On Monday, she uses a mixed unseen set and records the reason for each difficult decision. A cardiology question required extensive elimination, a pharmacology question exposed a missing mechanism, and a basic-science question was answered correctly through recognition of familiar wording. Those observations imply different next steps even before an algorithm makes a recommendation.
On Tuesday, she rebuilds the pharmacology mechanism using a concise explanation, then closes it and explains the distinction unaided. On Wednesday, she practises cardiology decisions under a time limit while checking whether her delay comes from interpretation or unnecessary reconsideration. Thursday's session samples neglected basic-science areas rather than simply repeating the ones already labelled weak.
At the weekend, she returns to a previously missed concept in a different question and completes another mixed set. This sequence is a proposed learning plan, not a report that Medibuddy, Pastest or iatroX automatically produced those exact assignments.
An effective personalised system should make such a plan easier to run. It should not make Priya less aware of what she is learning or why.
Can weak-area targeting create blind spots?
It can create an unhelpful study pattern when the learner treats every recommendation as a complete curriculum plan. Imagine that most of Priya's recent questions concern cardiology. Her cardiology estimate is now based on considerable exposure, while an apparently satisfactory score elsewhere rests on only a few questions. Treating both as equally reliable could hide untested material.
Keep two views of progress: performance in attempted material and coverage of the intended syllabus. They answer different questions. A useful review asks which areas have been sampled, which have been revisited after a delay and which have only been studied in topic-labelled sessions.
Medibuddy's published option to move between adaptive and traditional practice is relevant to this problem. It gives the learner a documented route to filtered practice, although whether a particular account makes breadth easy to manage still requires inspection. Its product description explains that design as of 24 September 2026.
Do not solve the problem by abandoning adaptive learning altogether. Instead, reserve part of the week for mixed work and periodically inspect the syllabus areas that the recommendation stream has not recently visited.
What happens after a correct guess?
A correct answer can conceal a learning need. Before revealing feedback, record whether you could defend the chosen option and reject the nearest alternative. This can be done in a notebook when a platform does not provide an explicit confidence field.
For Priya, a useful entry might read: "Correct, but I selected the familiar mechanism and cannot explain why the alternative is wrong." That is more actionable than simply flagging an entire specialty. Her next task is a comparison, not another long summary of the topic.
A tutoring conversation should end with a return to independence. Ask the learner to restate the distinction, identify the changed detail in a new original scenario and answer before help appears. A friendly conversation that never checks independent understanding may feel personalised while leaving the original error unresolved.
This is a proposed evaluation standard for all three products. It does not establish that one currently handles uncertain correct answers better than the others.
How to evaluate access before buying
Use a small, deliberate sample rather than selecting only favourite topics. Include something you understand well, something you recently misunderstood and something you have not revised. Inspect the sequence from attempt to explanation, review queue and later return.
Record the setting you used, the evidence visible in the interface and any question you could not resolve. Do not manufacture an error to test whether the platform is clinically safe; use ordinary educational questions and assess the clarity of its feedback. Avoid uploading a competitor's proprietary questions into another service.
For an adaptive recommendation, ask whether the underlying topic and reason are intelligible. For a tutor, ask whether it responds to your reasoning rather than merely restating the published explanation. For a review queue, check whether you can distinguish due material from untouched curriculum.
A short demonstration can establish that a control exists. It cannot validate the bank's entire content, the reliability of its estimated mastery or a claimed pass-rate advantage.
Price the workflow you will use
As advertised on 24 September 2026, Medibuddy's Part 1 options included £25 for one month, £50 for three months, £75 for six months and £100 for twelve months. Pastest's corresponding public page listed £99.99 for three months, £149.99 for six months and £199.99 for twelve months. These are dated advertised packages, not an assessment that their contents are equivalent. Sources: Medibuddy and Pastest.
Per iatroX pricing, September 2026, the combined question-bank, Tutor, planner, simulation and CPD subscription is £99 upfront annually, equivalent to £8.25 monthly billed annually, or £29 monthly. A short Medibuddy purchase can therefore cost less than either iatroX option for the relevant period. The purchasing question is whether you need a particular function, not which company can list the most methods.
Choose by the change you need
Medibuddy deserves close inspection when your main difficulty is deciding what to revisit and its adaptive-plus-traditional workflow appeals. Pastest deserves inspection when you want its combination of question practice, textbook support and AI assistance within one product. iatroX is relevant when targeted Socratic follow-up and a connected practice plan address an identifiable gap.
Retain an existing subscription when it already supports those tasks and the real problem is inconsistent use. Buying a second dashboard is not the same as developing a better revision process.
Frequently asked questions
Does adaptive revision mean a bank predicts my MRCP result?
No. Personalised question selection and mastery estimates do not establish a validated conversion to the official examination score.
Does Pastest have AI features as well as Medibuddy?
Its Part 1 page described tutor mode and Ask Pastest when reviewed on 24 September 2026. Compare the specific functions rather than treating AI as a feature unique to one provider.
Should I use two adaptive question banks?
Only when the second adds something you can identify, such as fresh assessment material or a more useful explanation method. Otherwise, split progress records and competing review queues can make the plan harder to manage.