Artificial intelligence and sustainable consumption efficiency in Chinese agriculture: evidence from SDG 12

Introduction Agricultural artificial intelligence innovation offers new opportunities to transform agricultural production and improve resource-use efficiency, yet less is known about how efficiently technological innovations are converted into sustainable production and consumption outcomes. Methods Using panel data for 30 provincial-level regions in China from 2019 to 2024, this study develops a dynamic three-stage network DDF-DEA model comprising agricultural AI technological innovation, sustainable agricultural production, and sustainable agricultural consumption. The model estimates both overall and stage-specific efficiencies. Results The results show that the mean overall efficiency was 0.7284 during the study period and increased with fluctuations, leaving substantial room for improvement. Regional efficiency was highest in eastern China, followed by western and central China, while provincial disparities remained evident. The mean efficiencies of the three stages were 0.4177, 0.9053, and 0.8880, respectively. Stage 1 was the main efficiency bottleneck in most provinces, whereas inefficiencies in Stages 2 and 3 were concentrated in a smaller number of provinces, indicating varying degrees of imbalance across the three stages. The mean Dagum Gini coefficient was 0.1097, and transvariation density accounted for an average of 46.56% of the overall disparity, suggesting that the overlap between regional efficiency distributions was an important source of inequality. The main findings remained broadly stable after indicator substitution, changes in moving-window width, the exclusion of municipalities, alternative stage weights, and the use of a slacks-based measure data envelopment analysis (SBM-DEA) model. Discussion These findings suggest that efficiency improvement strategies should be tailored to provincial stage-specific bottlenecks, with particular attention to strengthening agricultural AI innovation capacity and improving the conversion of technological innovations into sustainable production outcomes and of production outcomes into sustainable consumption outcomes.