The work aims to enhance the innovative application of generative adversarial network in the non-heritage landscape guide system, and then strengthen the Yao local cultural communication attributes. The elements of the guide system and the characteristics of Yao local culture were classified and summarized, and then the weights were assigned to the design elements and stylistic features of the guide system by the fuzzy hierarchy analysis method (FAHP). Then, a mapping matrix was constructed from the higher rated design elements and style features of the guide system, and finally, the selected style features were integrated into the Style Generative Adversarial Network (style-GAN) to assist the design of the Yao landscape guide system. A multi-disciplinary interdisciplinary cross-theoretical model was integrated, the FAHP was fused with style-GAN, and the guide interaction system of local cultural industries was reconstructed. With Yao embroidery as a case study, the organic combination of its traditional patterns and artificial intelligence technology can effectively promote the inheritance of local culture. This study also provides an innovative vision for the integration design of cultural guide system and related semantic refinement.