Does AI have utility in medical student surgical education? A comparative analysis of chatbots in answering standardized surgical multiple-choice questions

Abstract Purpose Artificial intelligence (AI) chatbots have potential as adjunctive medical education tools. AI chatbots can provide question-specific explanations with more supplemental content for students to learn from self-assessments. Chatbot performance on general surgery exams has not been studied at the medical student level. This study aims to assess the accuracy of popular low-cost chatbots—ChatGPT, Gemini, and Claude—in answering National Board of Medical Examiners (NBME) surgery practice questions for use in medical student education. Character count, as a proxy for question complexity, was assessed in relation to accuracy. Methods ChatGPT-4o mini, ChatGPT o3-mini, Gemini 2.0 Flash, and Claude 3.5 Sonnet were prompted to answer 20 multiple-choice questions from NBME Surgery Sample Items and provide justification on three attempts. Character count, answer choice, and explanation were recorded for each question. A logistic regression model assessed the relationship between accuracy and question character count. Results ChatGPT o3-mini and Claude 3.5 Sonnet scored 100% on all three attempts. Gemini 2.0 Flash scored 95% on all three attempts with an odds ratio of 0.904 [0.446, 1.831] ( p = 0.7794). ChatGPT-4o mini scored 95%, averaged on three attempts with odds ratios of 0.998 [0.993, 1.003] ( p = 0.4669). There was no statistically significant relationship between character count and accuracy. Conclusions The lack of correlation between question length and response accuracy implies that question complexity may not impact performance of these models. ChatGPT o3-mini and Claude 3.5 Sonnet outperform their counterparts on standardized general surgery exam questions, showcasing potential as supplementary tools for surgery students. ChatGPT-4o mini and Gemini 2.0 Flash have room for improvement to concurrently serve this purpose. Future models can continue to familiarize themselves with core surgical concepts to provide more comprehensive explanations.

Authors

Institutions

Publication Details

Journal
Global Surgical Education - Journal of the Association for Surgical Education
Published
2025-06-17
DOI
https://doi.org/10.1007/s44186-025-00369-3
Primary Topic
Artificial Intelligence in Healthcare and Education
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Does AI have utility in medical student surgical education? A comparative analysis of chatbots in answering standardized surgical multiple-choice questions

Angelo Cadiente, Burton Surick, C Implicito, Nílson Becker et al.
Global Surgical Education - Journal of the Association for Surgical Education
Artificial Intelligence in Healthcare and Education
article

Does AI have utility in medical student surgical education? A comparative analysis of chatbots in answering standardized surgical multiple-choice questions

Angelo Cadiente, Burton Surick, C Implicito, Nílson Becker, Natalia Dafonte
article en

Abstract

Abstract Purpose Artificial intelligence (AI) chatbots have potential as adjunctive medical education tools. AI chatbots can provide question-specific explanations with more supplemental content for students to learn from self-assessments. Chatbot performance on general surgery exams has not been studied at the medical student level. This study aims to assess the accuracy of popular low-cost chatbots—ChatGPT, Gemini, and Claude—in answering National Board of Medical Examiners (NBME) surgery practice questions for use in medical student education. Character count, as a proxy for question complexity, was assessed in relation to accuracy. Methods ChatGPT-4o mini, ChatGPT o3-mini, Gemini 2.0 Flash, and Claude 3.5 Sonnet were prompted to answer 20 multiple-choice questions from NBME Surgery Sample Items and provide justification on three attempts. Character count, answer choice, and explanation were recorded for each question. A logistic regression model assessed the relationship between accuracy and question character count. Results ChatGPT o3-mini and Claude 3.5 Sonnet scored 100% on all three attempts. Gemini 2.0 Flash scored 95% on all three attempts with an odds ratio of 0.904 [0.446, 1.831] ( p = 0.7794). ChatGPT-4o mini scored 95%, averaged on three attempts with odds ratios of 0.998 [0.993, 1.003] ( p = 0.4669). There was no statistically significant relationship between character count and accuracy. Conclusions The lack of correlation between question length and response accuracy implies that question complexity may not impact performance of these models. ChatGPT o3-mini and Claude 3.5 Sonnet outperform their counterparts on standardized general surgery exam questions, showcasing potential as supplementary tools for surgery students. ChatGPT-4o mini and Gemini 2.0 Flash have room for improvement to concurrently serve this purpose. Future models can continue to familiarize themselves with core surgical concepts to provide more comprehensive explanations.

Global Surgical Education - Journal of the Association for Surgical EducationVol. 4(1)
Hackensack University Medical Center (US), Center for Discovery (US)
Openalex Percentile: Top 91%
Artificial Intelligence in Healthcare and Education
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.