In this study, we assessed the interactive effects of phosphorus (P) application and irrigation methods on the abundances of marker genes (cbbL, cbbM, accA and aclB) of CO2-fixing autotrophs. We conducted rice-microcosm experiments using a P-limited paddy soil, with and without the addition of P fertiliser (P-treated-pot (P) versus control pot (CK)), and using two irrigation methods, namely alternate wetting and drying (AWD) and continuous flooding (CF). The abundances of bacterial 16S rRNA, archaeal 16S rRNA, cbbL, cbbM, accA and aclB genes in the rhizosphere soil (RS) and bulk soil (BS) were quantified. The application of P significantly altered the soil properties and stimulated the abundances of Bacteria, Archaea and CO2-fixation genes under CF treatment, but negatively influenced the abundances of Bacteria and marker genes of CO2-fixing autotrophs in BS soils under AWD treatment. The response of CO2-fixing autotrophs to P fertiliser depended on the irrigation management method. The redundancy analysis revealed that 54% of the variation in the functional marker gene abundances could be explained by the irrigation method, P fertiliser and the Olsen-P content; however, the rhizosphere effect did not have any significant influence. P fertiliser application under CF was more beneficial in improving the abundance of CO2-fixing autotrophs compared to the AWD treatment; thus, it is an ideal irrigation management method to increase soil carbon fixation.