This comparison is really about two meanings of personalisation, and naming them dissolves most of the confusion. Neural Consult personalises to your materials: on its public description, it builds summaries, flashcards, questions, study sessions and simulations around the lectures and documents you upload, "teach me from my institution's content". iatroX personalises to your performance: exam-specific question banks, adaptive targeting of the errors you make against a blueprint, spaced repetition and Socratic follow-up, "adapt to how I am doing against the examination". These are different machines answering different anxieties, my school's material versus my exam's standard, and iatroX is ours, so this piece is structured to make the choice legible rather than to win it; where curriculum-upload flexibility is the principal need, Neural Consult is the better fit, and saying so plainly is the point of the format.
What the upload model does well, and its structural risks
The strengths are real and specific: perfect topical alignment with what your faculty actually taught, in their emphasis and vocabulary; conversion of passive materials, slides, transcripts, notes, into active formats, questions, cards, sessions; and coverage of local or niche content no commercial bank maps, the regional formulary lecture, the school-specific framework. The structural risks are the mirror image, and they are generation risks, not vendor accusations: questions generated from your materials inherit your materials' errors and staleness; single-best-answer craft, one defensible answer, functioning distractors, is hard to guarantee at generation time, the validation-pipeline problem in consumer form, /blog/ai-draft-to-exam-ready-question-validation-pipeline; and blueprint coverage is bounded by what you uploaded, so the gap between your lecture set and the examination map is invisible precisely because the system never sees it. Upload-based study is strongest as a comprehension and consolidation layer on trusted materials, and weakest as the sole calibration for a national examination.
What the blueprint model does well, and its limits
The blueprint bank's strengths: coverage mapped to the examination rather than to any school's lecture list, so the unvisited domain becomes visible; curated items with review, corrections routes and stable difficulty intent; performance modelling across a defined map, which is what makes adaptive targeting and readiness reading meaningful; and continuity across examinations, the same learner history carrying from UKMLA through postgraduate banks. The limits, stated with equal plainness: a commercial bank does not contain your school's local content or emphasis; its explanations answer the blueprint's version of a topic, not your lecturer's; and for pre-clinical comprehension of genuinely new material, a question bank is a testing layer, not a first-teaching layer. The two models' weaknesses are, conveniently, each other's strengths, which is why the framing of rivals misleads.
The honest decision guide
Choose the upload model as your primary tool when the binding constraint is your institution's material: heavy local content, unusual curriculum structure, faculty-specific assessment, or a need to actively process a large volume of provided documents; Neural Consult's public feature set is built for exactly that user. Choose the blueprint model as your primary tool when the binding constraint is an external examination: UKMLA, USMLE, MCCQE, AMC or the postgraduate ladder, where coverage against the official map, curated item quality and delayed retention decide outcomes, which is the job iatroX is built around. Most students near examinations sensibly run both shapes in sequence rather than parallel, upload-based processing while a block is taught, blueprint-based retrieval as the examination approaches, with the switchover governed by one question: is my current risk misunderstanding the material, or being uncalibrated against the exam? Generated-versus-curated question quality remains the standing caveat on one side, and local-content blindness on the other; a student who can name which risk they are currently running has already chosen correctly.
Frequently asked questions
Can uploaded-material questions replace a curated bank for finals?
As the sole preparation, the coverage and calibration risks argue no; as the comprehension layer beneath a blueprint bank, they combine well, which is the sequencing above.
Does iatroX accept uploaded curriculum content?
Personalisation in iatroX runs on performance against exam blueprints rather than on document upload; that is an architectural choice, and students whose principal need is upload-based processing should weight the other model accordingly.
What about data and IP in uploaded lectures?
Upload rights on institutional materials vary and are worth checking before bulk upload anywhere; the what-not-to-paste rules for medical students apply to teaching materials as well as to patients.
Could the two models merge eventually?
The architectures are converging category-wide, upload layers adding curation, banks adding flexible ingestion; until a product genuinely does both well, the sequencing advice stands, and claims of doing both deserve the generated-question quality checks either way.
Which model suits graduate-entry and accelerated programmes?
Compressed timelines strengthen the blueprint side earlier: with less slack between teaching and examination, calibration risk dominates comprehension risk sooner, and the switchover question should be asked from month one rather than deferred to finals year.