Distractor handling is a backup skill. Concept recall is the job.

Distractor handling is a property of the answer choices: process of elimination, recognizing that an option with 'always' or 'never' is usually wrong, looking for the option that uses the same vocabulary the stem used, picking the most-detailed option when in doubt. Concept recall is a property of the source material: knowing what the slide said, cold, before you read the option list. Both are real skills. Only one of them survives the rewording NBME does between practice questions and real exam items.

They are not literally rewording UWorld stems, because UWorld is a third-party QBank and NBME writes its own items. The mechanism is the same in effect: NBME item writers are trained on the published item-writing guide, which explicitly avoids the patterns that test-tactic guides teach you to look for. Absolute language, grammatical mismatches between stem and answer, and longest-option-is-correct are all flagged for revision before items ship. So the distractor-elimination tactics that work on lazy MCQs do not work on items that passed NBME's review pipeline.

On the second pass you remember the wording, not the slide. Your brain optimizes for the cue. By pass three the question is a recognition test in disguise. UWorld has the strongest distractors of any commercial QBank, but it does not reword its own stems on revisit. So the moment you have seen a question once, the distractor work has already been done for you, and you are no longer training the recall mode that the real exam actually tests. The fix is to drill from a tool that rewords the stem and rotates the distractor pool every revisit.

When a card returns to a session, an LLM pass rewrites the stem (different opening words, different sentence shape, sometimes a switch from a direct question to a clinical scenario), and the distractor pool is rotated so the right-answer index moves around. The internal topic-pin (what the card is testing) stays fixed, which is what the spaced-repetition scheduler tracks. By revisit five you have answered the same underlying fact in five different surface forms. If you actually learned the slide, you still get it right. If you were memorizing the question, this is the take where it stops working.

There is a quality gate that runs at revisit time, the same rubric that runs on initial generation: factual correctness, clarity, distractor quality, question-type coverage. A reworded stem that drops below the gate gets rolled back to the prior version. The gate is the same one the held-out eval uses, where Studyly scores 81.3 versus Turbolearn 57.8. Methodology and full numbers are on the quality page.

When you genuinely do not know the fact. There is always going to be a fraction of items where the slide was not in your deck, the textbook chapter was assigned but never read, or the rotation question is something you have seen once. For those, narrowing four options to two with elimination, then guessing, beats blind guessing. The mistake is treating that fallback strategy as the primary mode. Drill concept recall as the default, keep distractor elimination in the back pocket for the items where you have no better option.

Both, but the leverage is different. Step 1 has a higher fact-density per item, so a reworded stem on the same fact closes the gap between recognition and recall faster. Step 2 CK has more clinical-scenario stems where the same fact gets dressed in a different patient presentation each time. The auto-rephrase pass produces both surface variants from one source, so you drill them both from the same upload.

Studyly is built around your professor's actual material. There is no public USMLE question bank inside Studyly itself; you upload your slide deck, First Aid chapter, lecture notes, or YouTube lecture, and the questions are generated against your source. If you want a public USMLE bank with first-pass items written by editors, UWorld and AMBOSS are the standard. The right combination is to drill UWorld for first-exposure breadth and to drill Studyly on your own decks for depth, where the auto-rephrasing keeps you out of recognition mode.

NBME self-assessments are the closest thing to a real-exam-grade item set you can practice on. Take them as scheduled checkpoints (every two to three weeks during dedicated), not as a daily QBank. They burn fast: there are only so many of them, and once you have taken one the items are spoiled. Save them for calibration, drill the rest of the time on items that are designed to come back with different surface forms.

The mechanism is identical, but the source material problem is bigger. Many IMG study tracks rely on First Aid plus UWorld plus a video series, with no slide deck from a course professor. The unique value of Studyly here is converting whatever long-form source you do have (a textbook chapter, a lecture transcript, your typed UWorld notes) into a per-source question pool that auto-rephrases on revisit. The five-minute-a-night habit is the part that keeps an eight-month dedicated period from collapsing into a final-month panic.