Abstract: Live streaming selling provides a new direct sales channel for numerous rural e-commerce sellers. Supporting live streaming selling for agricultural products has also become a new model for digital empowerment of rural revitalization. Based on Signaling Theory, this study utilizes a unique panel data from a large e-commerce platform in China to quantify the sales effects and heterogeneous impacts of live streaming selling on agricultural products by matching the live streaming adoption behavior of platform merchants with agricultural product sales data. The empirical model employs a state-of-the-art machine learning method, the causal forest model, to control for the self-selection bias of sellers choosing to live stream certain agricultural products. The core innovations of this study are: (1) It is the first to systematically reveal the causal effects of live streaming on agricultural product sales and their boundary conditions; (2) It innovatively identifies key moderating variables such as regional authenticity of agricultural products, product category differences (ornamental vs. edible), and streamer types (self-operated vs. third-party); (3) It employs the cutting-edge machine learning method—causal forest model, which effectively controls for the self-selection bias of sellers choosing specific agricultural products for live streaming, demonstrating stronger robustness compared to traditional methods. The study finds that, compared to traditional e-commerce, live streaming selling has a significant positive impact on both the sales volume and revenue of agricultural products. The positive impact of live streaming selling on agricultural product sales exhibits significant heterogeneity across different types of agricultural products and streamers. The positive impact of live streaming selling on sales is greater for agricultural products with higher prior sales, regionally authentic products, ornamental agricultural products (compared to edible agricultural products), and for streamers with more followers and self-operated streamers compared to third-party streamers. This study provides important managerial insights for rural e-commerce sellers, e-commerce platforms and policymakers.