Land degradation threatens agricultural productivity, but the specific microbial mechanisms linking degradation to crop yield loss remain unclear. This study investigated soil properties, bacterial communities, and maize yield across land degradation levels: nondegraded grassland (ND), lightly, moderately, and severely degraded cropland (LD, MD, and SD, respectively) in China's Songnen Sandy Land. Results showed that soil organic carbon, nitrogen, and moisture content declined significantly with increasing land degradation (p < 0.05). Land degradation significantly altered bacterial composition and reduced diversity. With intensifying land degradation, the abundance of sensitive taxa (e.g., Chloroflexi) declined, while that of stress-tolerant Arthrobacter increased. SD showed decreases of 9.76 % and 27.38 % in the Shannon and Chao indices, respectively, compared with ND (p < 0.05). Co-occurrence network analysis revealed a significant shift from predominantly positive interactions (synergy) in ND to increased negative correlations (competition) in degraded soils. Structural equation modeling revealed that land degradation directly impairs soil physicochemical properties and indirectly mediates maize yield losses by reducing bacterial diversity and network structure. The abundance of predicted microbial functional genes emerged as the most significant direct predictor of maize yield (beta = 0.729). The total contribution of bacterial communities (58.6 %) substantially surpassed the direct effect of soil physicochemical properties alone. Our findings suggest that changes in bacterial communities and microbial functional genes constitute key mechanisms driving maize yield loss in degraded soils. Restoring soil microbial diversity and functionality, therefore, offers an effective strategy to enhance agricultural productivity on degraded lands.