Abstract: Generative artificial intelligence (GAI) is a technology that produces text, images, sounds, and other content based on algorithmic models, and it has become deeply integrated into the field of education. Currently, primary school Chinese reading instruction encounters numerous challenges, including teaching homogenization, static texts, and empirical assessment methods. These issues hinder the enhancement of teaching quality as well as the growth and development of students. In light of this, the ACT-R architecture can be beveraged to model learners’ cognitive maps, construct a differentiated task system that outlines personalized reading pathways. Additionally, by integrating cross-media narratives to reconstruct multimodal texts, we can activate cognitive tension while defining ethical boundaries for cultural context transfer—thereby deepening text comprehension and fostering cultural inheritance. Furthermore, employing digital twins of historical contexts allows us to create embodied experiences. This approach facilitates the establishment of GAI companion cognitive scaffolds that promote meaning internalization and support critical thinking development. By collecting micro-evidence chains alongside flexible intervention systems for self-evolution, we can achieve closed-loop tracking and assessment of quality cultivation—ultimately supporting effective teaching implementation and realizing core Chinese literacy objectives.