As a non-destructive method, the proton nuclear magnetic resonance (H-1 NMR) technique with low echo time (TE, e.g., 0.07 ms) has been increasingly used for characterizing full pore size distribution (PSD) of shales. However, hydrogen contained in some components of the shale skeleton (e.g., kerogen and structural water) also can be detected by NMR in the case of low TE, resulting in a questionable PSD derived directly from the T-2 spectra of oil-saturated shale. In this study, eight shale samples with different organic and mineralogical components from the Jiyang Depression, China were investigated with regular NMR, low-temperature nitrogen adsorption (LTNA), NMR cryoporometry (NMRC), and mercury injection capillary pressure (MICP) techniques to propose a corrected NMR approach for characterizing shale PSD by considering the influence of the shale skeleton signals. The NMR relaxation characteristics (e.g., T-2 spectra and T-1 - T-2 map) of as-received shale, solvent extracted and dried shale (EX), and oil-saturated shale (OS) were discussed to reveal the NMR response from the shale skeleton itself at T-2 below 1 ms on the T-2 spectra. With the new approach, the NMR T-2 spectra of oil occurring in the OS shale were first obtained through inversion of the differentiated T-2 decay curves between the T-2 decay curves of the EX shale and OS shale and were then converted to PSD by combination of LTNA, NMRC, and MICP results. For pores with T-2 less than 1 ms, the PSD obtained from NMR T-2 spectra of the oil signals only compared well with the results of LTNA and NMRC, with a relative error of less than 15% in pore volume. In contrast, the relative errors of PSD obtained directly from the NMR T-2 spectra of oil-saturated shales were up to 134%. It was found that the higher total organic carbon shale contained, the larger errors in the PSD profiles, pore volume, and porosity that were calculated directly from the oil-saturated shale's NMR T-2 spectra. Compared with the traditional NMR methods, the corrected approach can provide a more accurate PSD for shales, especially for those organic-rich ones.