Artificial intelligence in mathematics education: a systematic review of pedagogical impact, cognitive outcomes, and ethical implications

Artificial intelligence (AI) is reshaping mathematics education, yet reviews remain fragmented by tool, level, or outcome. Following PRISMA 2020, we searched Scopus, Web of Science, and ERIC (2015–2025) and included 46 of 4,475 records, all published from 2022 onward and appraised with the Mixed Methods Appraisal Tool (κ = 0.762). Generative AI accounted for half of the studies. AI use was associated mainly with procedural, engagement, and self-regulation gains, whereas support for deep reasoning and proof remained fragile. Teachers acted as designers, supervisors, co-creators, and AI literacy mentors; ethical risks included hallucination, academic integrity, over-reliance, and a Global South gap. We propose a Pedagogical–Cognitive–Ethical (PCE) framework in which teachers mediate these interdependent dimensions.