# Comparing AI-Assisted Problem-Solving Ability With Internet Search Engine and e-Books in Medical Students With Variable Prior Subject Knowledge: Cross-Sectional Study

## TL;DR

This study compares how medical students with different levels of pharmacology knowledge perform when using AI, search engines, e-books, or self-knowledge to answer multiple-choice questions.

## Contribution

The study reveals that AI-LLM GPTs significantly boost problem-solving performance, especially for students with limited prior knowledge.

## Key findings

- Learned students outperformed naive students across all methods, with the largest effect size in AI-LLM GPT.
- AI-LLM GPT outperformed other methods for both groups, with naive students using AI scoring higher than learned students using Google or e-books.
- AI-LLM GPTs can enhance MCQ performance for students with limited prior knowledge, potentially transforming medical education.

## Abstract

Artificial intelligence (AI), particularly large language models (LLMs) such as ChatGPT (OpenAI), is rapidly influencing medical education. Its effectiveness for students with varying levels of prior knowledge remains underexplored.

This study aimed to evaluate the performance of medical students with and without formal pharmacology knowledge when using AI-LLM GPTs, internet search engines, e-books, or self-knowledge to solve multiple-choice questions (MCQs).

A cross-sectional study was conducted at a tertiary care teaching hospital with 100 medical students, divided into a “naive” group (n=50; no pharmacology training) and a “learned” group (n=50; completed pharmacology training). The study was started after approval from the Institutional Ethics Committee of Jawaharlal Nehru Medical College Hospital, Aligarh Muslim University (1018/IEC/23/8/23). Each participant answered 4 sets of 20 MCQs using self-knowledge, e-books, Google, or ChatGPT-4o. Scores were compared using analysis of covariance with self-knowledge scores as a covariate.

Learned students significantly outperformed naive students across all methods (P<.001), with the largest effect size in the AI-LLM GPT set (partial η²=0.328). For both groups, the performance hierarchy was AI-LLM GPT > internet search engine > self-knowledge ≈ e-books. Notably, the naive students who used AI scored higher (mean 13.24, SD 3.31) than the learned students who used Google (mean 12.14, SD 2.01; P=.01) or e-books (mean 10.22, SD 3.12; P<.001).

AI-LLM GPTs can significantly enhance problem-solving performance in MCQ-based assessments, particularly for students with limited prior knowledge, even allowing them to outperform knowledgeable peers using traditional digital resources. This underscores the potential of AI to transform learning support in medical education, although its impact on deep learning and critical thinking requires further investigation.