In response to the complexity of comprehensive evaluation of existing landscape design schemes and the non-linear relationship between influencing factors, this article designs a universal evaluation mode for various landscape design schemes using artificial intelligence algorithms. Firstly, the basic principles and related technologies for the evaluation of current landscape design schemes are analyzed, and an objective evaluation index system is established and the basic weights based on expert evaluation are also given. Furthermore, an evaluation model and algorithm based on self-organizing competitive neural networks are proposed and illustrated. Then, the raw data required for the evaluation indicators of the design scheme is collected, and a neural network is introduced to characterize the variation pattern between the design scheme level and the evaluation indicators. Finally, Matlab is used for empirical analysis of the case to explore the rationality and effectiveness of the evaluation model theory and algorithm. The experimental results show that the model has strong self-learning ability and it can perform nonlinear mapping without any simplification or assumption. In practical applications, the evaluation error is less than 5%, and it has characteristics such as high accuracy and short time consumption