Nuclear magnetic resonance (NMR) is used widely to characterize petrophysical properties of siliciclastic and carbonate rocks but rarely to study those of mixed siliciclastic–carbonate rocks. In this study, 13 different core samples and eight acidified core samples selected amongst those 13 from the Paleogene Shahejie Formation in Southern Laizhouwan Sag, Bohai Bay Basin, were tested by scanning electron microscopy (SEM), micro-nano-computed tomography (CT), and NMR. SEM and CT results revealed a complex pore structure diversity, pore distribution, and pore-throat connectivity in mixed reservoirs. Sixteen groups of NMR experiments addressed changes in these properties and permeabilities of mixed siliciclastic–carbonate rocks before and after acidification to determine its effects on such reservoirs. NMR experimental results showed no “diffusion coupling” effect in mixed siliciclastic–carbonate rocks. Distributions of NMR T2 cutoff values (T2C) are closely related to the pore structure and lithologic characteristics before and after acidification. The T2C index separates irreducible and movable fluids in porous rocks and is a key factor in permeability prediction. Centrifugation experiments showed that, before acidification, the T2C of mixed siliciclastic–carbonate rocks with 60–90% siliciclastic content (MSR) ranged widely from 1.5 to 9.8 ms; the T2C of mixed siliciclastic–carbonate rocks with 60–90% carbonate content (MCR) ranged from 1.8 to 5.6 ms. After acidification, the T2C of MSR ranged widely from 2.6 to 11.6 ms, the T2C of MCR ranged from 1.5 to 5.6 ms, and no significant difference was observed between MCR reservoirs. Based on an analysis of the morphology of NMR T2 spectra, we propose a new T2 cutoff value prediction method for mixed siliciclastic–carbonate rocks based on a normal distribution function to predict various T2C values from morphological differences in NMR T2 spectra and to calculate the irreducible water saturation (Swir), i.e., the ratio of irreducible total fluid volume to effective porosity. The reliability of the proposed method is verified by comparing predicted T2C and Swir values with those from NMR experimental results. New experiments and modeling demonstrate the applicability of NMR for the petrophysical characterization of mixed siliciclastic–carbonate rock reservoirs. Our results have potential applications for identification and evaluation of mixed siliciclastic–carbonate rock reservoirs using NMR logging.