AIGC has become increasingly embedded in textile education, promoting an automatic and advanced teaching process. However, its autonomous production of textile-related content often exhibits reconstruction bias and poses teaching risks due to limitations in training data and the inherent complexity of textile systems. Therefore, this study adopts a focused review of representative literature examines how such bias occurs in three core domains of textile knowledge from textile material knowledge, textile machinery knowledge, and textile process knowledge aspects. The result shows that AIGC frequently misrepresents fiber properties and yarn structures, conflates machinery components and operational constraints, and oversimplifies multi-stage textile processes into inaccurate or incomplete workflows. These bias create substantial teaching risks, including the formation of flawed conceptual models, disruption of skill acquisition, and long-term misconceptions that may hinder learners’ professional competence. By identifying the mechanisms underlying AIGC-induced reconstruction bias and its teaching risks, this study provides guidance for the critical and informed integration of generative AI into textile education.