Artificial intelligence teaching assistants: a scalable solution for supporting struggling medical students

PURPOSE: Large language models (LLMs), such as OpenAI's ChatGPT, have demonstrated tutoring benefits in small-scale pilot studies within focused areas of medical education. This study evaluated the large-scale implementation of AI-teaching assistants (AI-TAs) within a compulsory medical school course. METHOD: A quasi-experimental observational study with a mixed-methods was conducted to assess the impact of AI-TAs in a compulsory first-year medical school course at the University of Toronto in 2024. The research team developed AI-TAs using OpenAI's ChatGPT-4o and introduced them as a supplementary resource at the course's midpoint. They analyzed exam performance among the students who used AI-TAs (n = 87) and students who did not (n = 206). Additionally, surveys (n = 18) and interviews (n = 10) explored student perceptions of AI-TAs' effectiveness, usability, and impact on learning. RESULTS: Students who would later adopt AI-TAs had significantly lower pre-intervention exam scores than their peers (83.8% vs 88.1%; t(118.8)=-3.82, P<.001). They also more often failed to meet the course's assessment standard pre-intervention (24.1% vs 6.4%). After AI-TA adoption, performance converged for both early users (86.1% vs 86.5%; t(95.63)=-0.35, P=.72) and late adopters (85.6% vs 86.5%; t(26.08)=-0.52, P=.61), with similar proportions falling below the course's standard (4.4-6.4%). Thematic analysis and surveys identified three key advantages of AI-TAs: (1) Reliable and accurate educational support, (2) Efficient application across learning activities, and (3) Improved psychological safety, promoting engagement in active learning. CONCLUSIONS: AI-TAs correlated with improved exam performance, fewer students in academic difficulty, and enhanced student engagement through a psychologically safe learning environment. These findings suggest that AI-TAs can serve as a scalable, cost-effective tool to support struggling students, while complementing traditional instruction.

Authors

Institutions

Publication Details

Journal
Academic Medicine
Published
2026-07-02
DOI
https://doi.org/10.1093/acamed/wvag208
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

Artificial intelligence teaching assistants: a scalable solution for supporting struggling medical students

Chris Gilchrist, Alina Sami, Mark Adkins, F. Y. Leung et al.
Academic Medicine
Artificial Intelligence in Healthcare and Education
article

Artificial intelligence teaching assistants: a scalable solution for supporting struggling medical students

Chris Gilchrist, Alina Sami, Mark Adkins, F. Y. Leung, Anne McLeod
article en

Abstract

PURPOSE: Large language models (LLMs), such as OpenAI's ChatGPT, have demonstrated tutoring benefits in small-scale pilot studies within focused areas of medical education. This study evaluated the large-scale implementation of AI-teaching assistants (AI-TAs) within a compulsory medical school course. METHOD: A quasi-experimental observational study with a mixed-methods was conducted to assess the impact of AI-TAs in a compulsory first-year medical school course at the University of Toronto in 2024. The research team developed AI-TAs using OpenAI's ChatGPT-4o and introduced them as a supplementary resource at the course's midpoint. They analyzed exam performance among the students who used AI-TAs (n = 87) and students who did not (n = 206). Additionally, surveys (n = 18) and interviews (n = 10) explored student perceptions of AI-TAs' effectiveness, usability, and impact on learning. RESULTS: Students who would later adopt AI-TAs had significantly lower pre-intervention exam scores than their peers (83.8% vs 88.1%; t(118.8)=-3.82, P<.001). They also more often failed to meet the course's assessment standard pre-intervention (24.1% vs 6.4%). After AI-TA adoption, performance converged for both early users (86.1% vs 86.5%; t(95.63)=-0.35, P=.72) and late adopters (85.6% vs 86.5%; t(26.08)=-0.52, P=.61), with similar proportions falling below the course's standard (4.4-6.4%). Thematic analysis and surveys identified three key advantages of AI-TAs: (1) Reliable and accurate educational support, (2) Efficient application across learning activities, and (3) Improved psychological safety, promoting engagement in active learning. CONCLUSIONS: AI-TAs correlated with improved exam performance, fewer students in academic difficulty, and enhanced student engagement through a psychologically safe learning environment. These findings suggest that AI-TAs can serve as a scalable, cost-effective tool to support struggling students, while complementing traditional instruction.

Academic Medicine
University of Toronto (CA)
Quality Education
Openalex Percentile: Top 11%
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.

Artificial intelligence teaching assistants: a scalable solution for supporting struggling medical students — Chris Gilchrist, Alina Sami, et al. · Academic Medicine (2026) | TGRS Research Map | TGRS