With the rapid development of information technology, the emergence of Generative AI and cross-modal learning technology provides essential opportunities for innovative teaching in landscape architecture. This teaching research aims to enhance the effectiveness of cognitive teaching in Chinese classical gardens and explore and construct a new model of garden cognitive teaching empowered by AI. This teaching utilizes landscape painting restoration as a medium and designs a “five-step progressive” teaching framework to guide students through a complete learning loop, from knowledge acquisition to visual interpretation, AI practice, and critical reflection. Through a phased AI technology process, guided landscape scene restoration is achieved. This technological path aims to organically integrate garden knowledge, garden painting images, and cutting-edge AI technology, and build a complete technical support from data processing, model construction, to result generation. At the level of teaching implementation, this framework and technical process revolve around representative landscape painting restoration tasks. Through a combination of theoretical teaching and practical technical operations, students are guided to collaborate in groups to complete the entire process from literature research to AI restoration. The teaching results indicate that this model has preliminarily achieved AI restoration of garden painting. Students not only generate garden images with high realism and historical atmosphere but also deepen their understanding of classical garden design principles, spatial imagery, and cultural connotations. Compared with traditional methods, the teaching model that integrates AI has shown significant advantages in stimulating students’ initiative and inquiry, achieving dynamic and experiential cognitive modes, enhancing multimodal and efficient knowledge acquisition, and systematically cultivating digital technology applications and critical thinking. It effectively shortens the cognitive cycle and provides a practical basis and methodological reference for the paradigm transformation of landscape education from “knowledge imparting” to “ability-oriented and inquiry learning”. It has positive significance for promoting innovation and sustainable development of landscape education in the AI era.