[Purpose/significance] Cross-modality is one of the key attributes of generative artificial intelligence. This study employs a social semiotics perspective to conduct a micro-analysis of user-AI cross-modal interaction process⁃es, revealing the necessity and urgency of multimodal literacy education for users within the framework of AI literacy education. [Method/process] This study adopts mixed methods research, using questionnaires, participatory observa⁃tion, and multimodal discourse analysis to collectively demonstrate the significant roles of literary literacy, aesthetic lit⁃eracy, and other types of literacy in user-AI cross-modal interactions. [Result/conclusion] The current framework and content of AI literacy education should be deepened from four aspects: multimodal comprehension ability, cross-modal transformation ability, multimodal creation ability, and multimodal judgment ability；user-AI interactions exhibit the
characteristic of“transliteracy”, and creative interactions between users and AI require a broader and more interrelated set of literacy；the paradigm of AI literacy education should shift from“skills trainer”to“literacy architect”, the core competitiveness of future libraries will be reflected in the ability to integrate multiple types of literacy in educational practices.