PM2.5 satellite remote sensing is the most powerful way to acquire the PM2.5 distribution and variation at a large scale with high resolution. Thus, PM2.5 remote sensing methods have been widely developed and applied in multiple environmentally related research areas in recent decades. Hence, the purpose of this review is to summarize these methods, required input data and main applications of PM2.5 and its remote sensing components. In general, two-step methods have been used for estimating PM2.5, which first retrieves the aerosol optical depth (AOD) and estimates PM2.5 from the AOD with other supplemental data containing the temporal or spatial variation impact on PM2.5 or data correlated with PM2.5 variation by different AOD-PM2.5 models. The AOD-PM2.5 models have been developed by using different methods, including empirical-statistical models (single or combined statistical models and big data-based machine learning methods), CTM-based models and semi-empirical/physical models. Current research can provide high-resolution (e.g. daily variations at 1 km and hourly variations at ~1 km) PM2.5 products, which have been widely used in air pollution management, health impact assessments, numerical data assimilation and climate impact analyses. This review summarizes the current research on method development, application, achievement and remaining challenges in remote sensing of PM2.5 and its components, which are essential for further improvement of the methods and accuracy of PM2.5 remote sensing and are likely applicable to other PM2.5 component remote sensing methods in the future.