With rapid development of urbanization, the atmospheric pollution caused by PM_2.5 has attracted widespread concern. Land as the underlying surface of atmosphere, which has a significant effect on PM_2.5. This study aimed to explore the spatiotemporal variation of PM_2.5 and its relationship to land use in China during 2005–2017, then put forward suggestions for land use optimization to improve air quality. The results suggested that PM_2.5 generally showed a downward trend, with obvious spatial pattern. High-pollution areas were formed in the North China Plain, the Middle-Lower Yangtze River Plain, the Sichuan Basin and the Taklimakan Desert. Furthermore, PM_2.5 was significantly positive correlated with the area of cultivated land and artificial surfaces, and cultivated land fragmentation was conducive to the decline of PM_2.5, while contiguous expansion of artificial surfaces would increase PM_2.5. Forest and grassland had an inhibitory effect on PM_2.5, and the larger area, the more complex shape, the better PM_2.5 retention effect. Moreover, cultivated land and forest were the main land use types that affect PM_2.5, and the effect of cultivated land was greater than that of forest. The percentage of cultivated land landscape, the shape index of forest and water bodies were the main landscape metrics that affect PM_2.5, and their effect on PM_2.5 decreased in turn. Therefore, strengthening the pollution control of cultivated land, promoting the land recuperation, further expanding the scale of forest and grassland, enriching vegetation types, promoting the grain for green, reasonably controlling the scale of urban development, are conducive to reduce PM_2.5.