This study aims to investigate the potential of generative artificial intelligence (AI) technology in the generation and innovation of Mongolian embroidery patterns. It seeks to address the limitations of traditional embroidery pattern design, which is often time- consuming and inefficient, and to enhance the market competitiveness of Mongolian embroidery products. A substantial corpus of Mongolian embroidery patterns has been assembled, and image processing software has been employed to enhance the processing. A generative adversarial network (GAN) and a variational self-encoder model are constructed to learn and train Mongolian embroidery patterns, with the objective of generating new patterns that exhibit a combination of traditional style and modern design concepts. The experimental results demonstrate that the Mongolian embroidery patterns generated by AIGC technology retain the defining characteristics of traditional patterns in terms of form, while also exhibiting a greater degree of diversification in design style through the creative generation of the model.