Carbon emissions are critical to sustainable development, with land use and land cover change (LUCC) being a key influencing factor. However, clarifying the variations in land-use carbon emissions driven by multiple factors remains challenging. As an ecological pilot in China’s Yangtze River Delta (YRD), the YRD Demonstration Zone provides an ideal case for studying land-use carbon dynamics. This study analyzed the evolution characteristics and drivers of land-use carbon emissions in the zone using Landsat imagery from 1990 to 2025 and machine learning algorithms. The land-use classification model achieved an accuracy of 0.93, with transferable accuracy averaging 0.91. Notable transitions included persistent conversion of paddy fields to terrestrial vegetation and impervious surfaces, while wetlands remained stable. Carbon accounting revealed increases in both sinks and sources, with net emissions remaining source-dominated. Using the LMDI (Logarithmic Mean Divisia Index) model and collinearity analysis, main drivers of net emissions were identified and ranked as: economic development > carbon sink pressure > industrial-ecological balance > regional population > economic structure. This study clarifies how land-use transitions affect regional carbon balance, offering a scientific basis for low-carbon development and sustainable land management in the YRD.