Answers are grounded in this paper and cite their sources.
Artificial intelligence (AI) is entering medical school admissions and residency selection, the two most consequential selection points in U.S. physician training, amid high application volume, pass/fail USMLE Step 1 reporting, and renewed fairness scrutiny. This structured narrative review synthesizes research and policy, legal, and accreditation sources on AI-assisted screening, predictive analytics, narrative and video analysis, applicant-facing tools, and applicants’ use of generative AI. Early single-institution studies show that it is feasible for machine learning models to approximate faculty screening, reduce workload, and occasionally surface overlooked applicants. These studies establish feasibility, not effectiveness or generalizability. Most models are validated against proximal outcomes such as interview invitations, subgroup bias is inconsistently assessed, and one simulation found demographic bias in large language model selection decisions despite equal qualifications. We propose the Door–Gate Matrix, an author-developed conceptual framework classifying each AI use by whether it expands or restricts applicant opportunity and whether applicants can see and contest it. Each quadrant carries distinct governance obligations, and four diagnostic questions guide review. The central risk is the silent gate, a restrictive use applicants cannot see or challenge. We outline how the untested framework could be validated and how it handles model updates and more capable AI.
No takes yet. Share an insight, caveat, or question.
Traylor et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: