Abstract: Aerosols, which offset a third of the greenhouse gas forcing, remain the primary source of uncertainty in climate monitoring. Satellite products, models, or reanalyses provide time series of Aerosol Optical Depth (AOD), each with distinct strengths and weaknesses. This study evaluates the temporal stability of these datasets from 2003 to 2022 using spatially representative long-term AERONET measurements as a reference. AOD has decreased globally since 2015, driven by anthropogenic emissions reduction in Europe, the US, and particularly in East Asia. Aerosols continue to rise in India due to growing anthropogenic emissions, and in South America due to a shift from declining to increasing organic matter aerosol trends. While all products capture these regional trends, they diverge in terms of magnitude, seasonal variability, and temporal patterns. Only CAMS EAC4 reanalysis reproduces the AOD decrease observed at AERONET stations in Europe and the US with a trend in the bias below 5%/decade. The other products evaluated show drifts above 10%/decade, underestimating the AOD decrease in these regions. MERRA-2 and the C3S multi-satellite products exhibit spurious jumps likely caused by methodological changes. Satellite-based (MODIS and MISR) AOD trends have a positive drift compared to AERONET, potentially due to the use of static aerosol composition datasets. The low satellite revisit time and a calibration drift could also contribute to the MISR spurious trend. An analysis of the reduced temporal sampling of satellites revealed that the impact at the station level is small, suggesting that the drifts found can be attributed to the products. Our results suggest that the aerosol Essential Climate Variable could be better served by reanalysis rather than by direct satellite observations.