Weed control and biodiversity conservation are required for sustainable agricultural production. The impact of long-term chemical fertilization on the agroecological environment has attracted widespread attention. To validate the effects of machine-transplanted rice using side-deep fertilization (MRSF) on weed control and weed community composition in paddy fields, a field experiment was conducted in a double-cropping rice system at the Dongting Lake in southern China. Seven fertilization treatments for early- and late-season rice were tested, including control (CK), conventional fertilization (CF), and MRSF with controlled-release blending fertilizer (MRSF-CRBF) at various nitrogen (N) rates (MC, 10% MC, 20% MC, 30% MC, and 40% MC). The highest yields of early- and late-season rice were obtained with 10% MC and 40% MC, respectively, with an increase by 2.5% and 5.0% compared to the CF treatment. The CK treatment exhibited the least weed species, while CF treatment the highest Shannon index across growth stages. The highest weed biomass of early- and late-season rice was observed under CF treatment, which was 4.5–37.7% and 4.5–24.1% higher than those in the 0–40% MC treatments, respectively. Soil pH, available potassium and available phosphorus (P) in the double-cropping paddy field were negatively correlated with weed biomass. Soil available N content was positively correlated with weed biomass in early- versus late-season rice. Soil available P content indirectly affected total weed biomass via pH and available N. Redundancy analysis showed that weed biomass was negatively correlated with soil physicochemical properties, and soil available N and organic matter content had significant effects on weed distribution. In conclusion, the MRSF-CRBF strategy effectively maintained suitable available N and P content during the late-season of rice growth, controlled weed growth, and achieved a beneficial dynamic balance between weed species and total biomass.