Improving CO2 flooding in deep oil reservoirs requires a mechanistic understanding of how chemical additives regulate oil-CO2 interactions across scales. In this study, theoretical calculations, molecular dynamics (MD), and computational fluid dynamics (CFD) were integrated to investigate the microscopic mechanisms and concentration-optimization strategy of chemical-assisted CO2 flooding. A multiscale model was established to account for the effects of chemicals on CO2 dissolution, extraction, and diffusion in oil, thereby linking molecular interactions with microscopic seepage behavior. The main results are as follows: (1) a newly defined miscibility-degree index showed that Span-series surfactants, particularly 0.5 wt% Span 20, significantly enhanced CO2-oil miscibility; (2) the effects of chemicals on CO2 mass transfer were concentration dependent, with 0.1 wt% Span 20 increasing CO2 solubility in oil by 20.64% and the diffusion coefficient by 12.38%, whereas extraction efficiency increased by 44.60% at 0.5 wt%; (3) chemicals improved oil recovery through coupled mechanisms, including reduced oil-wall interaction energy (with the optimum effect at 0.1 wt% Span 20) and strengthened molecular interactions between CO2 and oil; (4) pore type influenced recovery efficiency, with Span-series chemicals markedly improving recovery from small and blind-end pores, whereas anionic surfactant sodium bis(2-ethylhexyl) sulfosuccinate (AOT) mainly enhanced recovery from small pores; and (5) 0.2 wt% Span 20 was identified as the optimal concentration, while adsorption loss requires the practical injection concentration to be adjusted to 0.54%–0.68%. These findings provide a theoretical basis and methodological support for designing chemical-assisted CO2 flooding schemes for deep reservoirs.