Artificial Intelligence (AI) is one of the core technologies of the newround industrial revolution in the 21st century.It has been comprehensively affecting the development of the economy and society.Based on the industrial application of AI and the theoretical model of Acemoglu and Restrepo (2020),this paper takes robotics as the measurement of AI to examine AI’s effects and mechanisms on agricultural productivity,using cross-country data of 1991 to 2018 from the World Bank Database and the IFR Database.The empirical results show AI has significantly positive effects on agricultural productivity measured by value added per worker.From a mechanism perspective,AI significantly substitutes agricultural labor,but it has no significant impacts on industrial structure yet.Interestingly addition,as for the heterogeneity of effects on agriculture productivity of countries in different developing stages,AI has remarkably positive effects on high-income countries,while there exist negative effects on the middle and low-income countries.To develop modern agriculture,we should systemically strengthen the development and application of AI in agriculture,further facilitate agricultural labor transfer and boost agricultural productivity.