Abstract: Investigating the temporal and spatial evolution of public perception of air quality, along with its driving factors, is crucial for fostering rational assessments of air pollution risks and enhancing public ecological well-being, sense of gain, satisfaction, and security with governmental environmental protection efforts. Utilizing provincial-level data from 2003 to 2021 in China, this study employs spatial autocorrelation and spatial regression models to analyze the spatial heterogeneity of public air quality perception while examining its driving factors through the lenses of physical space, social space, and information space. The findings reveal: (1) An overall improvement in public air quality perception across China is evident; however, significant spatial disparities exist. The trend indicates a hierarchy: ‘Western > Central > Northeastern > Eastern’. (2) Analysis of various factors′ spatial effects shows that, within physical space, local air pollution levels negatively impact improvements in public perception of air quality. In contrast, within social space, economic development positively correlates with enhancements in local perceptions, while local environmental protection initiatives also benefit neighboring areas’ perceptions. Regarding information space, both internet development levels and the government environmental information disclosure index (PITI) demonstrate significance only under certain direct or indirect conditions. (3) Overall assessment indicates that advancements in local environmental protection efforts and PITI positively contribute to improving public air quality perception; while elevated levels of local air pollution and economic growth inhibit such improvements. Finally, this paper advocates for policy recommendations, including optimizing the factors influencing public perceptions, namely physical space, social space, and information dissemination strategies. It also suggests further refining the spatial resolution for measuring perceived air quality data, increasing the density of monitoring data related to actual air quality, and exploring synergistic governance models that address both objective and subjective aspects of air pollution management.