Shelf and NBME Score Reports¶
The Prompts This Guide Uses
Score Report Study Planner
Turns an exam performance report into an honest, interleaved study plan.
Have ready: your report as text, your next exam date and honest weekly study hours, the resources you actually have, and your own three takeaways, written first.
Study Plan Progress Check
Compares two reports and says honestly whether the plan is working.
Works with any capable assistant in the Tool Directory.
The Five-Minute Version
The Workflow¶
The loop, closed at both ends
Your score report
INSIGHTS or in-house, de-identified
Your own read
three takeaways, written first
Planner prompt
an honest, interleaved plan
Sanity check
real hours, every claim traced
Your advisor
the plan becomes the agenda
Next assessment
keep, shrink, or change the plan
The next assessment produces the next report, and the loop starts again.
The plan is a hypothesis; the next assessment is the test.
- Get your data out. Open INSIGHTS, the National Board of Medical Examiners (NBME) dashboard (what it holds is below), download the score report or export the tables, or collect your in-house report. Paste it into a text file and strip your name, student number, and any exam identifiers before any artificial intelligence (AI) tool sees it. The planner needs none of them, and stripping them takes a minute.
- Write your own read. Three takeaways, before any model sees anything. Two minutes that make everything after them work better (why).
- Run the Score Report Study Planner. It will ask for your report, your timeline, your honest hours (after classes, work, and everything else: honest, not aspirational), your resources (question bank, lecture notes, review materials, so the plan names real things), your own read, and your test-taking patterns, then produce an honest assessment, a deficit map, the quick wins (biostatistics, epidemiology, ethics, and communication, checked explicitly when your exam tests them, as Step 1 and the comprehensive exams built to its outline do), an interleaved week-by-week plan, test-taking drills for any strategy problems it finds, three questions for your advisor, and what to re-measure on your next assessment.
- Sanity-check the plan. Does it start with the flagged areas that make up a large part of the exam, as NBME's own advice does? Are the weekly hours ones you actually have? Does every week keep your strong areas warm rather than parking a month on one subject? Does every claim about your performance trace to something in your report?
- Take it to your advisor. The plan's advisor questions are the agenda. Advisors see patterns no model can: how this exam fits your trajectory, what worked for students in your exact position, and when the problem is not the studying at all.
- Close the loop. After your next assessment, run the Study Plan Progress Check with the old plan and the new report. It will say honestly which of three things happened: the plan worked, the plan was not followed, or the plan was followed and did not work, and each has a different next move.
Good Practice for This Task¶
- A classmate's report is theirs to share. If a classmate wants help, send them this page so they can run the planner on their own report, or sit beside them while they do. The AI Responsible Use Policy is the reference for student information.
- Performance data, not questions. The report's content-area descriptions are all the planner needs. NBME exam questions are copyrighted and confidential, so leave out any you remember from the exam; to dig into a concept you missed, work from your question bank or your notes instead.
- Treat single-exam areas as hypotheses. Content-area scores on one exam are less precise than the total score, as NBME's reports point out. A pattern across two or more exams, or a flag that matches your own sense of weakness, is evidence; one dip is a lead to investigate.
- A missed plan is information. A plan you did not follow says something about the plan's size, not about you, and the progress check treats it that way; the next plan should be one you can keep.
For advisors
Both prompts work in an advising meeting with the student driving on their own account and screen, which lets the student choose what to share and teaches the method at the same time. If you work with a student's report yourself, remove the name, student number, and exam identifiers first: the planner needs none of them. The planner's output ends with questions the data cannot answer; that section exists to make your meeting sharper, not to replace it.
About This Task¶
Turn an exam performance report (a comprehensive exam, a shelf exam, a self-assessment, or an in-house exam) into a study plan that is honest about where you stand, fits the hours you actually have, and gets tested against your next assessment instead of drifting. The report is not the hard part; many students receive more performance data than they ever use. The hard part is reading it without flinching, turning it into a plan that survives a bad week, and closing the loop.
What You Already Have¶
Before any AI enters the picture, know your data. If you have taken an NBME exam, you already have a dashboard: INSIGHTS, reached through the MyNBME examinee portal. It collects your NBME self-assessments, subject exams, and comprehensive exams from the past two years, plus any Customized Assessment Services exams your school runs through NBME, under the email address the exam was tied to; older score reports come from the school. Your United States Medical Licensing Examination score reports also appear in the exam list for about a year after they are posted, though not in the detailed views; students of schools outside the United States receive them first through the Federation of State Medical Boards (FSMB) portal. INSIGHTS has four tabs: your exam list with downloadable score reports, per-exam results, question-level detail, and comparisons across multiple takes of the same exam type.
The five report types, compared
The reports carry different rulers, so read yours against the right row:
| Report | Score scale | Pass estimate | What this report adds |
|---|---|---|---|
| Basic science comprehensive exam | Equated percent correct (EPC) | Probability of passing Step 1 if you had tested within a week (NBME's newest sample report adds a four-week estimate) | Each content area against Step 1 first-takers from US and Canadian medical schools, flagged lower, about the same, or higher, with up to six suggested areas of focus split between organ systems and disciplines |
| Clinical science comprehensive exam | 1 to 300, estimating Step 2 Clinical Knowledge performance | Probability of passing Step 2 Clinical Knowledge if you had tested within a week | The same content-area comparison and up to six areas of focus, against Step 2 Clinical Knowledge first-takers |
| Subject (shelf) exam | EPC | None; your school sets any pass mark | Each content area against first-time takers from US and Canadian medical schools at the end of the same clerkship or course, up to five suggested areas of focus, and the national share answering each question correctly |
| NBME self-assessment | The comprehensive self-assessments use the matching comprehensive exam's scale | The same estimate as the matching comprehensive exam | Time spent per question, the only NBME report with pacing data, plus each question with its answer explanation; if timing is your suspected problem, this is where the evidence lives |
| In-house exam | Set by the course | None; the course sets any pass mark | Category and question-level reports through the exam platform; the same method above applies to them |
On both comprehensive exams, read the pass estimate beside the likely score range the report draws against the Step 1 or Step 2 Clinical Knowledge low-pass range: NBME advises extra caution when the two overlap.
Most tables in INSIGHTS export to a spreadsheet. For an in-house block exam, the Study Schedule Builder is the closer fit: it plans day by day from your syllabus and teaching calendar, where the planner above works week by week toward NBME exams.
The shape of an INSIGHTS exam-results view
My ExamsExam ResultsQuestion DetailsResults Comparison
Total scoreequated percent correct (EPC); the basic science comprehensive exam adds a probability of passing Step 1
Content areas, your EPC vs a national comparison group
Suggested areas of focus: the dashboard lists up to six, split between systems and disciplines.
An illustration of the layout, not a real report; NBME's own demo linked below shows the live version.
See it for real before your first exam: NBME publishes an interactive demo of INSIGHTS and an official user guide that walks every tab.
Read It Yourself First¶
Before you run any prompt, open your report and write down your own three takeaways: where you think you stand, what you think went wrong, and what you would change. This is not a ritual. Reading your own performance data is a skill you will need on every future exam, every licensing step, and eventually on the audit of your own practice, and you do not build it by outsourcing the first read. Then run the prompt and compare. Where the AI's read matches yours, plan with confidence. Where it disagrees, that disagreement is the most useful thing either of you produced, and it is exactly what to bring to your advisor.
Where AI Helps, and Where It Hurts¶
AI is strong at the layer most students skip: translating a wall of flags and percentages into an ordered plan, sizing that plan to the hours you actually have, keeping strong areas in rotation while you repair weak ones, and asking the test-taking questions (Did you change answers? Did you run out of time?) that separate strategy problems from content problems. For an advisor, it applies the same rigor to the twentieth report of the week as to the first.
It fails in predictable ways. Models overread noisy data: a single content area on a single exam is less precise than the total score, as NBME's reports point out, and an AI will happily build you a month of cardiology around what may be statistical noise. They invent numbers: percentiles, national averages, and pass probabilities that are not in your report. They flatter: reassurance is cheap for a model and expensive for you, because false comfort costs an exam attempt. And a model cannot know you. Your advisor has seen many students walk this path and can hear what you are not saying; the model sees one pasted report. The prompts in this guide are built to resist the first three failures. The fourth is why the loop ends at your advisor's door, not the model's.
Related: AI and the Residency Application for the year that follows, and Module 6: Clinical Contexts for the wards.
Content last reviewed October 12, 2026. Review dates are maintained as described on the About page.