As generative artificial intelligence (GAI) becomes increasingly embedded in educational practice, the inherent limitations of traditional self-regulated learning theory are becoming more pronounced: ontologically, it neglects the mediating role of the body; epistemologically, it adheres to a linear regulation paradigm; and in practice, it faces the challenge of fragmented technological support—making it difficult to effectively address learners’ complex developmental needs in learning to learn. With its generativity, human-computer interactivity, higher-order capabilities, and autonomy, GAI transcends the dualistic cognitive structure of subject and object. It co-constructs a triadic system of “subject-mediator (technology)-object” with the learner, thereby reshaping the meaning of “learning to learn” toward a model characterized by human-machine collaboration and directly addressing “Jobs question”. In this context, generative self-regulated learning (GSRL) emerges. Guided by dialectical constructivism and Vygotsky’s cognitive development theory at the macro level, and supported by micro-theoretical frameworks of human-technology relationship theory, embodied cognition theory, and dialogic theory, GSRL constructs a triple-loop pathway—“pure thinking-embodied experience-deep cognition”. This pathway not only redefines the logical framework of SRL but also enriches the contemporary connotation of “learning to learn”, signifying a genuine revolution of learning. At the same time, GSRL may pose risks such as loss of reflexivity, alienation of practice, deconstruction of subjectivity, and disruption of intersubjective communication. Therefore, a dialectical balance must be maintained between the instrumental rationality of technology and the value rationality of education, ensuring that education remains centered on human development and contributes to the construction of a Chinese educational knowledge system with both contemporary relevance and indigenous significance.