Methodology

What each figure rests on

Counted from the same file the estimate reads, every time this page is built. 204 schools.

Read the last column as the honest limit of this product. A range marked Estimate is arithmetic off the median, not that school's real distribution, and is labelled that way wherever it appears.

National outcome grids

Every estimate is built on the AAMC's Table A-23, the official grid of applicants and acceptees to U.S. MD programs, aggregated across the 2023-24 through 2025-26 cycles: 155,905 applicant outcomes, 45.1% of whom were accepted somewhere. Rates in that table mean acceptance to at least one U.S. MD school. We parse the AAMC's published workbook directly rather than transcribing it, so a refresh is one command and cannot drift by hand.

DO figures come from the AACOM (AACOMAS) grid for 2018-19 through 2020-21. Two honest caveats we refuse to paper over: AACOM publishes matriculation rates, not offer rates, so DO cells describe who enrolled rather than who was accepted; and that grid is older than the MD one because AACOM has not released a newer edition at this granularity.

Per-school data

204 schools: 160 MD and 44 DO programs across 49 states, 34,130 first-year seats. Six of them were accredited and enrolling but missing from this site entirely until the official tables turned them up: Roseman and UT Tyler on the MD side, and Baptist Health Sciences, Meritus, Orlando and Rocky Vista Montana on the DO side.

Every school now carries per-field provenance: each number records the table it came from and whether that is an official source, the school’s own published figure, something computed from official counts, or still an estimate. The weak spots are visible rather than buried in one prose footnote.

Class size comes from AAMC FACTS Table A-1 for every MD school and AACOM’s per-COM report for every DO college recruiting through AACOMAS: 202 of the 204. MD schools also get the in-state share of their class and of their applications from Table A-1. DO colleges get their in-state split and their entering-class GPA and MCAT from their own AACOM Choose DO Explorer profile, which AACOM publishes but does not verify.

Acceptance rates used to be hand-set guesses, and it showed: 158 MD schools shared 42 distinct values between them and 40 DO colleges shared five. They are now computed from each school’s own published applications and matriculants, scaled by the roughly 50% yield medical schools report, for 201 of the 204. That also corrected a systematic error: the figures we had hand-checked for UCSF, UCLA, UC San Diego and Columbia were each exactly that school’s matriculants over its applications, which is a matriculation rate, not an acceptance rate, and about half of it.

What is still not a school’s own figure, and marked as such: the 25th-75th percentile ranges, which almost no school publishes. Where one does (NEOMED, UCLA), or publishes a 10th-90th range (Stanford), that is used. Everywhere else the band is the school’s median plus or minus the spread inside a typical MD class, about 0.13 of a GPA point and 3.5 MCAT points. That spread is measured, not guessed: AAMC Table A-16 gives the standard deviation of every U.S. MD matriculant in 2025-26 (GPA 0.21, MCAT 6.5), and taking out the part explained by schools differing from each other, which their own published class averages measure, leaves the spread within a class. It matches NEOMED’s published range to within a few hundredths. It is still a typical class, not that school’s, and the per-field provenance says so on every school.

Class GPA and MCAT figures come from 124 schools' own websites, read page by page, with the page and the passage kept for every value.

The AAMC's free, public FACTS tables (A-1, A-16 and A-23) are the backbone of every estimate. What we do not use is MSAR, the AAMC's paid school guide, because its terms prohibit reuse in other products. So every per-school figure here comes from sources published openly: AAMC and AACOM tables and the schools' own websites.

National benchmarks

The applicant-versus-matriculant table uses AAMC FACTS Table A-16 for MD (2025-26 matriculants: 3.81 total GPA, 3.75 science GPA, 512.1 MCAT, across 23,440 matriculants) and AACOM's applicant and matriculant reports for DO (2024: 3.59 total GPA, 3.49 science GPA, 502.97 MCAT).

The experience-hour rows in that table are different, and we label them differently. Neither the AAMC nor AACOM publishes matriculant experience hours, so there is no national figure to compare against: those rows use advising guidance and are marked as such rather than presented beside the GPA and MCAT rows as though equally sourced.

What the AAMC does publish about experiences is participation, from the Matriculating Student Questionnaire sent to every entering MD student. In 2023, 94.2% had shadowed a physician, 90.4% had volunteered in a healthcare setting, 87.6% outside one, and 52.9% had done a college lab research apprenticeship. That last figure is worth knowing: about half of matriculants did no formal college research.

Selection criteria, read by hand

For 198 of the 204 schools we read the school's own published mission statement and selection-criteria pages and wrote a structured profile: what the school says it weighs, the applicant it describes wanting, and the most common mismatch. That is 351 source citations across 226 distinct official domains and roughly 21,400 words of per-school analysis, completed August 2, 2026. The six schools added since are too new to have published a mission statement (Rocky Vista Montana’s own page has none), so they use weights derived from their NIH volume, mission tags and in-state share rather than an invented profile.

The per-school weightings are our reading of those published criteria, not an official formula; no school releases exact weights. Where a school publishes no differential emphasis, its entry records stated mission emphasis and says so. Where a school's own pages were unreachable and an entry leans partly on secondary admissions sources, its source note flags that individually.

Research funding and accreditation

Research intensity is checked against FY2024 funded-project counts from the official NIH RePORTER API for all 204 schools, matched by exact registered organization name so look-alike institutions are not conflated: a token query for New York University otherwise matches every State University of New York campus. A school whose query returns no funded projects is recorded as a genuine zero, which is real information about a college that does little federally funded research; one whose query returns rows but no exact match is recorded as unmatched rather than guessed.

Accreditation is now covered for both accreditors. LCME status for MD programs was retrieved August 2, 2026 and verified school by school. DO programs are accredited by COCA, not LCME, and were previously left unlabeled rather than defaulted to a status we could not verify. That gap is closed from COCA’s own May 2026 directory, which puts four of the DO colleges here in pre-accreditation: approved to enrol students but not yet through the full sequence, typically without a graduating class. Any school not labeled holds full standing with its own accreditor.

How the estimate is computed

Every estimate starts from the school’s own acceptance rate and is adjusted by how applicants with your GPA and MCAT have actually fared, from the AAMC’s Table A-23. Because that table is binned, a figure between two brackets is interpolated between the published cells on either side rather than snapped to one, and never extrapolated past the outermost bracket.

Table A-23 is the only published table of acceptances by GPA and MCAT, so it supplies the shape of the estimate for DO schools as well as MD. AACOM’s DO grid counts matriculations instead, and at the top it runs backwards (the strongest DO applicants are admitted and then enrol at MD schools), so read as acceptances it rated a 4.00/524 applicant below a 3.55/503 one. Each DO school’s own acceptance rate and admitted median supply its level; the DO grid is still what a DO applicant sees in the national comparison, labelled as matriculation.

Science GPA is placed on the total-GPA scale before it is used, since the grid and every school median are in total GPA. Science GPA runs about a tenth of a point lower and more spread out for the same people, so averaging the raw numbers penalised a completely ordinary profile. Once spread is accounted for, AAMC Table A-16 shows the two separate matriculants from applicants by the same amount, so they are weighted equally.

A school’s own admitted median corrects for its applicants being stronger than the national pool, damped by a square root. That exponent was chosen by checking the model against reality: summing each school’s estimate back up, across the official 19.5 applications per applicant, and asking which values reproduce the published “accepted to at least one school” rates. 0.5 and 0.6 bracket the published rates, leaning less than a point high and low respectively, and full normalization fits clearly worse; we use 0.5, which also needs no shift in overall level. At schools whose admitted median sits below the average applicant nationally (Howard, Meharry and the Puerto Rico schools among them), the damping used to rate a typical admit below the school’s own acceptance rate, which cannot be right, and no longer does.

How well does it hold up? Run the same way against Table A-23, for the 89% of applicants in cells with at least a 5% acceptance rate, the model reproduces the published outcomes with a log-odds error of 0.07 and leans under a point high. Every GPA row is within one point of the AAMC figure, the weakest included: where the AAMC publishes 0.0% and 1.0%, the model gives 0.3% and 1.4%. Most of that accuracy came from using each school’s own published class GPA and MCAT: with the secondary compilation they replaced, whose GPAs ran about a tenth of a point low, the same check leaned nearly two points high. The check is published as a script in the codebase, so anyone changing the model can see whether they made it better or worse.

The residency adjustment needs two shares: how much of a school’s class is in-state, and how much of its applications are. Both are published per school in AAMC Table A-1, so for every MD school it runs on real numbers, and it self-cancels where residency does not matter. That second share is why one national figure was never going to work: it is 22% by application volume but runs from 0.5% at Vermont to 99.8% at Mercer, and two schools that both fill about 85% of their class in-state can be opposite propositions: 76% of applications to Texas A&M come from Texans, against 9% of applications to Indiana from Hoosiers. Where a school admits more than its own state as residents, its published rates are used instead: the University of Washington publishes a rate for each state it serves (15.5% for Washington, 25% for Montana, 27% for Idaho, 44% for Alaska and 50% for Wyoming) and 0.4% for everyone else, and Jefferson reserves twenty seats a year for Delaware. Before this, a Montanan and a Californian both read 4% at UW.

The arithmetic runs in odds rather than raw percentages, which is the standard way to carry a rate measured across all applicants onto one specific school. Bands are derived from the resulting estimate, so the label and the number can never disagree. Anything under 1% is shown as “<1%” rather than rounded up.

Every ranked row is computed in code from this data. The AI writes the narrative sections of the paid report; it does not write, and cannot alter, any number in the rankings.

What this cannot tell you

These are descriptive summaries of how applicants with numbers like yours fared historically, weighted by what schools say they value. They are not predictions of your outcome. Admissions committees read essays, letters, interviews and context this tool cannot see, and no estimate here has been calibrated against individual admission decisions. Treat every figure as a reference class, not a forecast.