Crustal motion observations from Global Positioning System (GPS) networks have not yet been fully exploited in previous studies on glacial isostatic adjustment (GIA) and mantle rheology structure. In this study, we have isolated GIA signals from vertical velocity observations at rigorously selected (over 2,000) GPS stations from the Nevada Geodetic Laboratory (NGL) by removing the effects of atmospheric and oceanic loading, as well as changes in the contemporary glaciers. We have also attempted to include hydrological and sea-level loading corrections based on the most updated model products, but found them still not accurate enough for GIA-related studies. Therefore, we recommend the GPS-derived global GIA uplift rate dataset MIDAS-AO without hydrological and sea-level loading corrections applied. Under the constraints of MIDAS-AO uplift rates, we refined the VM5a viscosity model and obtained two revised viscosity profiles, VM5aR_AO1 and VM5aR_AO2, that differ by an extra layer in the transition zone of the latter profile. With respect to VM5a, VM5aR_AO1 indicates a slight increase of viscosity within the upper mantle, while VM5aR_AO2 favors a softer upper part of the upper mantle and a stiffer transition zone. Maps of the variations of model-dataset misfits show that our new viscosity profiles commonly recover a better fit for sites located at the Scandinavian Peninsula and south of the Hudson Bay.