<![CDATA[Generative artificial intelligence (GenAI) has been rapidly adopted in education, yet its impact on students' metacognitive learning remains poorly understood. Most existing reviews focus broadly on learning outcomes or technology acceptance, neglecting specific metacognitive processes, planning, monitoring, evaluation, and reflection that underpin self-regulated learning. This systematic review synthesizes empirical evidence on how GenAI influences metacognitive learning, which pedagogical approaches support it, and the risks that emerge. Following PRISMA 2020 guidelines, a systematic search of Scopus, ERIC, and Google Scholar identified 12 empirical studies published between 2020 and 2026. The findings reveal that GenAI's effect on metacognition is fundamentally conditional: structured scaffolding improves self-regulation, whereas unrestricted use leads to metacognitive laziness, cognitive offloading, and a decline in self-regulation. Four effective pedagogical approaches were identified: metacognitive prompting, guided inquiry with scaffolding, reflective journaling, and feedback literacy instruction. Significant risks include passive acceptance of AI answers, illusion of competence, and unequal benefits for students with weaker feedback literacy. This review presents a Conditional Metacognitive Support framework to guide GenAI integration. For educators, effective implementation requires structured prompts, reflective activities, and explicit feedback literacy instruction.]]>
