Crop rotation systems profoundly influence soil phosphorus (P) dynamics through physicochemical and microbial interactions. The mechanisms regulating P availability under various rotational practices remain poorly understood. This five-year field experiment investigated the effects of four rotation systems (WM: wheat–maize; WP: wheat–peanut; WS: wheat–soybean; MV: maize–hairy vetch) on soil P fractions, phosphatase activities, P-cycling gene abundance, and their interactions with soil properties. The WM rotation substantially reduced soil pH (6.29) while increasing labile P fractions (Ca₂-P) and moderately labile P (Al-P, Fe-P, and Ca₈-P), which was attributed to enhanced acid phosphatase activity. The WP rotation elevated soil pH (8.13) but reduced P availability due to calcium–P immobilization. The MV rotation stimulated microbial P cycling, exhibiting the highest phoD (2.01 × 10⁶ copies g⁻¹) and phnK (33,140 copies g⁻¹) gene abundance, which was linked to green manure-induced microbial activation. Redundancy analysis identified soil pH, total nitrogen, and stoichiometric ratios (C/N and N/P) as key shared drivers of P fractions and enzymatic activity. Partial least squares path modeling (PLS–PM) indicated that crop rotation directly regulated P availability through pH modulation (r = −0.559 ***) and the C/N ratio (r = 0.343 ***) while indirectly regulating P fractions through phosphatase activity. Lower C/N ratios (<10) across all rotation regimes amplified the carbon limitation in the process of P transformation, indicating that exogenous carbon inputs and appropriate stoichiometry in the soil should be optimized. The results of this study inform the selection of suitable crop rotation patterns for sustainable agriculture.