Physical activity during recess can provide an important opportunity for secondary school students to accumulate health-enhancing movement within the school day. However, in high-density urban schools, limited campus land and uneven spatial conditions may constrain students’ physical activity during recess. Although previous studies examined the role of the school environment in shaping students’ physical activity, little is known about how different types of campus public spaces and students’ perception of such spaces are associated with recess physical activity. In this study, such associations are examined in three high-density urban schools in Guangzhou, China among 900 students in grades 10–12. The students’ moderate-to-vigorous physical activity (MVPA) during recess is measured by using Huawei Band 8 wearable devices, and their primary activity spaces and perception of their spatial environment are determined by using a structured questionnaire. The campus public spaces are classified as sports field, courtyard, plaza, undercroft space and corridors, and the students’ perception of their environment is assessed across four dimensions: usability, accessibility, safety and comfort. Descriptive statistics, Pearson correlation analysis and multiple linear regression are used for the data analysis. Results show differences in recess MVPA levels across the three schools, with high activity levels observed in the school with superior spatial resources and public space conditions, as well as the use of and MVPA behaviour in the different public spaces. Sports fields were generally associated with high use and higher MVPA levels, whereas corridors mainly supported students’ movement between destinations and brief resting and were associated with relatively low MVPA levels. Courtyards, plazas and undercroft spaces show varied patterns, in which activity is related to specific spatial conditions, such as the scale, openness, paving and shading conditions and facility availability. The perception analysis indicates that usability, comfort, accessibility and safety are significantly and positively associated with recess MVPA, with usability showing the strongest association. The regression model can explain 54.7% of the variance in recess MVPA levels. The findings suggest that in similar high-density urban secondary school contexts, spatial support for recess physical activity depends on not only the amount of available space but also the activity-supportive characteristics and perceived environmental quality of the campus public spaces. Improvement of the usability, comfort and accessibility of campus public spaces may provide favourable spatial conditions for students’ physical activity during recess.