Amidst the widespread application of virtual agents in online content production, our study explored the impact of virtual instructors’ attributes on users’ cognitive load in health short videos. Through a 2 (Formal Clothing Style vs. Casual Clothing Style) × 2 (Formal Linguistic Style vs. Casual Linguistic Style) × 2 (Treatment Theme vs. Lifestyle Theme) experimental design, we found that: (1) a casual clothing style and the theme related to healthy lifestyle contributes to lower extraneous cognitive load; (2) a casual linguistic style and the theme related to healthy lifestyle induce lower intrinsic cognitive load; (3) the interaction between clothing, linguistic and thematic styles did not exhibit a significant effect. Our research expands understanding of the mechanisms by which virtual images affect user cognition in health short video environments. Stakeholders can optimize the design and delivery of health short videos, ultimately enhancing their effectiveness in engaging users in health promotion efforts.