The pore system significantly affects methane accumulation and migration; therefore, accurate evaluation of pore compressibility is crucial for shale gas exploration and exploitation. However, the compressibility and preservation mechanisms of different size pores in deep shale gas reservoirs are still unknown. In this study, the compressibility and heterogeneity of different scale pores in the Wufeng–Longmaxi (WF–LM) Formation deep shale were evaluated by experimental porosity and nuclear magnetic resonance results under different confining pressures. Moreover, the influences of the material composition and pore structure on pore compressibility were analyzed, and the pore preservation mechanism of deep shale was revealed. The results show that deep shale has a weaker pore compressibility than shallow shale. With an increase in effective stress, pore stress sensitivity decreases exponentially, while pore compressibility first fluctuates and then gradually weakens. Stress sensitivity can be divided into sensitive, transitional, and stabilized stages. The stress sensitivity and compressibility of WF are the strongest, followed by those of LM114and LM111–3. Macropores are the easiest to compress, while micropores are the hardest to compress. The compressibility of mesopores is closest to that of the whole pore system. Pore heterogeneity increases gradually with pore compression. Adsorption space heterogeneity has the strongest stress sensitivity, followed by mesopore and pore system heterogeneity. The change in pore system heterogeneity is mainly caused by micropore compression. Stress initially mostly compresses pore space and then primarily reconstructs the pore structure. Pore compressibility is restricted by the micropore development degree, pore shape, and pore heterogeneity. Biogenic quartz and reservoir fluid overpressure are the keys to the pore preservation of deep shale. This study demonstrates that LM111–3benefit the most from the enrichment and development of shale gas, which provides a theoretical basis for the optimization of deep shale gas fracturing targets.