Nuclear magnetic resonance (NMR) logging has been widely used to characterize rock pore size distribution in recent years. However, theory and field applications suggest that data obtained from the NMR logging is not always valid in hydrocarbon bearing formations due to the effect of the bulk relaxation of hydrocarbon. In this study, we examined the effect of hydrocarbon to NMR T-2 spectrum, using the kerosene and transformer oil with different viscosity, based on lab NMR experiments. In the laboratory, a total of 20 core samples, drilled from three different types of formations, were applied for NMR experiments under four saturated conditions. For each core sample, the saturated conditions contain irreducible water saturation, fully saturated with brine, hydrocarbon bearing condition, and saturated with residual oil. The results of our study show that the bulk relaxation of hydrocarbon has little effect on the NMR T-2 spectra for conventional and tight sands. However, the T-2 spectra for medium to low permeability sands are highly affected. Accordingly, we proposed a method to remove the effect of hydrocarbon to NMR T-2 response. This method was based upon laboratory NMR measurements of fully brine saturation and residual oil condition for 10 core samples drilled from medium to low permeability sands. The reliability of this method was further evaluated by comparing the corrected results with the lab NMR measurements. Finally, we applied this method to field application in the Cretaceous of Shinan area of Junggar Basin, northwest China. We were able to successfully acquired T-2 distributions for full water saturation from field NMR logging, and they are well used for formation pore structure evaluation.