Electrochemical coupling between carbon and nitrogen species to generate high-value C-N products, including urea, presents significant economic and environmental potentials for addressing the energy crisis. However, this electrocatalysis process still suffers from limited mechanism understanding due to the complex reaction networks, which restricts the development of electrocatalysts beyond trial-and-error practices. In this work, we aim to improve the understanding of the C-N coupling mechanism. This goal was achieved by constructing the activity and selectivity landscape on 54 MXene surfaces by density functional theory (DFT) calculations. Our results show that the activity of the C-N coupling step is largely determined by the *CO adsorption strength (Ead-CO), while the selectivity relies more on the co-adsorption strength of *N and *CO (Ead-CO and Ead-N). Based on these findings, we propose that an ideal C-N coupling MXene catalyst should satisfy moderate *CO and stable *N adsorption. Through the machine learning-based approach, data-driven formulas for describing the relationship between Ead-CO and Ead-N with atomic physical chemistry features were further identified. Based on the identified formula, 162 MXene materials were screened without time-consuming DFT calculations. Several potential catalysts were predicted with good C-N coupling performance, such as Ta2W2C3. The candidate was then verified by DFT calculations. This study has incorporated machine learning methods for the first time to provide an efficient high-throughput screening method for selective C-N coupling electrocatalysts, which could be extended to a wider range of electrocatalytic reactions to facilitate green chemical production.