This study explores oil recovery mechanisms of static imbibition in a tight sandstone under different imbibition pressures, simultaneously optimizing imbibition agents. To this end, the static imbibition experiments of two common agents, polyacrylamide (PAM) slick water and anion–nonionic surfactants, are conducted under atmospheric and reservoir pressure (20 MPa). The interfacial tension and contact angle of these two imbibition fluids are also measured. Herein, the entire recovery period and imbibition equilibrium time vs. pressure are determined. Based on NMR and high-pressure mercury injection measurements, the contributions of pores with different sizes to the displacement recovery during imbibition are quantified. Under atmospheric pressure, the recovery rate with the surfactant was measured higher than that of the PAM slick water. The main reason was that the former has a lower interfacial tension (0.0961 mN/m), stronger hydrophilicity (average contact angle 27.7°), and stronger oil-displacing effect. Under a reservoir pressure of 20 MPa, the surfactant had lower recovery rate than the PAM slick water, while the latter enhanced the recovery further. Under atmospheric pressure, both agents recovered the crude oil in the medium-sized and larger pores whereas, under reservoir pressure (20 MPa), they mainly recovered oil from the smaller and medium-sized pores. At higher imbibition pressures, both agents recovered more oil from the smaller and medium-sized pores, and less from micropores and larger pores. This indicated that higher pressures can further improve the driving force of fluid replacement, to improve oil production from finer and medium-sized pores significantly. Under atmospheric pressure, both agents reached imbibition equilibrium in approximately 20 days while this period was reduced for the surfactant and slick water to 16 and 12 days, respectively. Based on the results PAM slick water is recommended for EOR purposes in tight sandstone which can be generalized to similar formations around the globe.