The paper is devoted to the application of AI methods for the analysis of plant communities based on remote sensing images. The analysis is carried out to select territories that can be included in the regional monitoring database of valuable steppe communities in the Samara Region. The analysis apparatus is an artificial neural network in the form of a multilayer perceptron used for pixel-by-pixel classification of remote sensing image series for several vegetation seasons. The paper proposes a way to aggregate the classification results for different seasons, taking into account the high confidence of belonging to classes. The feasibility of using relief in classification is investigated. To verify the results obtained by AI, ground-based surveys of the selected territories were carried out. These surveys confirmed the class affiliation of plant communities of a significant part of the selected areas, which proved the efficiency of the proposed approach and the feasibility of using AI methods to optimize field research.