Objective To explore the influence of artificial intelligence (AI)-assisted learning on the learning performance of medical students.

Methods A cross-sectional survey of 3 010 medical students from 4 medical colleges in Anhui province was conducted by using the general population sociological data questionnaire, AI use questionnaire, learning engagement scale and learning continuation scale.

Results Among the 3 010 medical students, 2 257 (74.98%) had used AI tools to assist learning. Compared with the students who had not used AI tools to assist learning, the students who had used AI tools to assist learning had higher levels of learning engagement and performance (P ＜ 0.01). Multiple linear regression analysis showed that, compared with medical students who had not used AI tools to assist learning, the use and persistence of AI tools to assist learning had a positive impact on the learning engagement and learning performance of medical students (P ＜ 0.01). Among medical students who insisted on using AI tools to assist learning, the use of AI tools to explore low-cognitive-closed (LC) and high-cognitive-closed (HC) problems had a positive impact on their learning engagement and learning performance (P ＜ 0.01). The propensity value matching method was used to test the robustness of the effect of AI tool use on learning engagement and learning performance, and found that, compared with the medical students who had not used AI tools to assist learning, the medical students who had used AI tools had significantly higher learning engagement (average treatment effect on the treated, ATT) = 0.209, P ＜ 0.01 and learning performance (ATT = 0.213, P ＜ 0.01); compared with the medical students who did not insist on using AI tools to assist learning, the medical students who insisted on using AI tools had higher learning engagement (ATT = 0.103, P ＜ 0.01) and higher learning performance (ATT = 0.081, P ＜ 0.01). The effect decomposition results of the OB decomposition method on the sources of the gaP between learning engagement and learning performance showed that, compared with medical students who did not use AI tools to assist learning, the average learning engagement of medical students who used AI tools to assist learning was 0.219 points higher, which could explain 92.24% of the difference in learning engagement of students between the two groups, and the average improvement in learning performance was 0.205 points higher, which could explain 90.73% of the difference in learning performance of students between the two groups.

Conclusions Medical students using AI tools to assist learning can effectively improve learning engagement and learning performance, and helP achieve learning goals.