Abstract: Under the context of urbanization, cropland fragmentation has emerged as a critical issue constraining agricultural modernization and large-scale development, posing a significant threat to food production security in China. This study employed a comprehensive evaluation model integrating the landscape pattern index, entropy method, and Pearson correlation coefficient test to assess the spatio-temporal evolution of cropland fragmentation at the district and county levels in Guangdong Province. Additionally, the land use transfer matrix was utilized to analyze the dynamic changes in cropland, while the variance inflation factor (VIF) test and geographic detector method were applied to identify the key drivers of cropland landscape fragmentation. The results show that: 1) Temporally, cropland fragmentation in Guangdong Province exhibited a trend of "slight decline — continuous increase — rapid decrease", with substantial fluctuations over time. Spatially, the cropland fragmentation index was highest in western Guangdong and lowest in northern Guangdong. Western Guangdong experienced the most significant changes, while northern Guangdong remained relatively stable, indicating pronounced spatial disparities in fragmentation. Over time, the fragmentation pattern evolved into a radial distribution extending outward from the Pearl River Delta, with the high-fragmentation zones shifting from western and eastern Guangdong to the Pearl River Delta. 2) The evolution of cropland fragmentation was significantly influenced by the gross output value of agriculture, forestry, animal husbandry, and fishery (GOV-AFAF); urbanization rate; GDP; altitude; slope; average annual precipitation; kernel density; and proximity to roads. Among these, socioeconomic factors—particularly GOV-AFAF, as well as urbanization rate and GDP—played a dominant role. GOV-AFAF was identified as the most critical factor, underscoring that socioeconomic dynamics are the primary inhibitors of cropland fragmentation in Guangdong Province and a key determinant of its spatial variability. 3) Four typical driving mechanisms were identified for the evolution of cropland landscape fragmentation in Guangdong Province: First, exemplified by Yunan County, the implementation of cropland-to-forest conversion programs reduced fragmented cropland patches on hillsides, promoting more consolidated cropland distribution. Second, as observed in Lianzhou City and Huangpu District, industrial restructuring and rapid socioeconomic development led to insufficient emphasis on cropland protection. Third, illustrated by Gaozhou City, enhanced policy interventions for cropland protection mitigated fragmentation. Fourth, typified by Yangchun City, adjustments in crop cultivation and land use transitions resulted in a net reduction of cropland area.