Search for a command to run...
This is an accepted article with a DOI pre-assigned that is not yet published.The integration of Artificial Intelligence (AI) into mathematics education has generated a polarized debate: while some studies present AI as a personalized tutoring tool, others warn of its effects on critical thinking and creativity. Both positions conceive learning as the acquisition of information, overlooking its historical, affective, and embodied dimensions. This study proposes a reframing of that debate through the Theory of Objectification (TO) developed by Radford (2021, 2023), a sociocultural framework that conceptualizes mathematical learning as an embodied, collectively enacted, and historically situated process. The objective is to reinterpret the role of AI in middle and high school mathematics education through the principles of the TO, with emphasis on implications for teaching practice. The methodology consists of a systematic literature review following the PRISMA protocol, with a corpus of 32 documents published between 2019 and 2025, analyzed through a theoretical-interpretive approach that contrasts the principles of embodiment, joint activity, and historicity against available empirical evidence on AI in mathematics classrooms. The findings show that AI does not think, it acts as a cultural sign that reorganizes mathematical activity, and its uncritical integration carries three identifiable pedagogical risks: the disembodiment of thought, the de-historicization of mathematical knowledge, and the erosion of joint activity. At the same time, the analysis demonstrates that a pedagogically situated integration can expand the semiotic horizon of the classroom without displacing the relational and embodied dimensions of learning. The study concludes with guidelines for designing activities in which AI functions as a semiotic provocation rather than a cognitive substitute.