Related Experiment Video
Updated: Jan 15, 2026

Evidence-based Knowledge Synthesis and Hypothesis Validation: Navigating Biomedical Knowledge Bases via Explainable AI and Agentic Systems
Published on: June 13, 2025
Death of the Personal Statement: Qualitative Comparison Between Human-Authored and Artificial Intelligence-Generated
Matthew J Vaccaro1, Ishpriya Sharma2, Andrea P Espina Rey1
1From the, University of Central Florida College of Medicine, Orlando, FL (Vaccaro, Espina-Rey, Lyman, Zhang).
Background:
Medical school admissions committees use personal statements as subjective measures of applicants' values, motivations, and experiences. Students at varying levels of education report that they would or do use artificial intelligence (AI) in their academic pursuits. It has not been investigated how effectively medical school admissions committees can distinguish between AI-generated and human-authored personal statements, or whether the use of AI provides students an advantage.
Study Design:
Human-authored personal statements were retrieved from the 2019 application cycle (before AI chatbots). ChatGPT 4.0 was used to generate personal statements from summaries of the human-authored essays. In a prospective, single-blind, randomized controlled trial, medical school application readers evaluated unique combinations of AI-generated and human-authored personal statements for essay quality (on a 7-point Likert scale), speculated authorship (human or AI), and speculation confidence (on a 5-point Likert scale). ZeroGPT, an AI-detection tool, also speculated authorship.
Results:
Seventeen medical school application readers completed 325 scoring rubrics across 309 essays. Readers demonstrated a 56% accuracy in correctly identifying authorship, whereas ZeroGPT showed 91% accuracy. Readers were more likely to assume human authorship of higher-scoring essays (Mann-Whitney U test, p < 0.001). AI-generated essays scored better than their human-authored counterparts (Wilcoxon signed-rank test, p = 0.020) with mean scores of 5.02 ± 1.21 and 4.67 ± 1.33, respectively.
Conclusions:
AI-generated personal statements were rated more highly and were nearly indistinguishable to medical school application readers. Despite ZeroGPT's higher accuracy in detecting AI use, its rate of false positives remains unacceptably high for use in medical school admissions. The role of personal statements in medical school admissions requires urgent reconsideration to maintain credibility.
Related Concept Videos
Non-equilibrium in the Cell
Reliability and Validity
The Sense of Self: Reflected Self-Appraisal and Social Comparison
Comparing Experimental Results: Student's t-Test
Comparing the Survival Analysis of Two or More Groups
What is Genetic Engineering?
