<![CDATA[Artificial Intelligence (AI) is increasingly integrated into mathematics education to support personalized, adaptive, and student-centered learning. However, evidence concerning its simultaneous contribution to students’ mathematical problem-solving skills and self-directed learning remains fragmented. This study aimed to systematically analyze publication trends, forms of AI implementation, contributions to mathematical problem solving and self-directed learning, and research gaps in AI-supported mathematics education. A Systematic Literature Review was conducted following the PRISMA 2020 guidelines. Relevant journal articles published between 2021 and 2025 were retrieved from Scopus, OpenAlex, and Google Scholar using predefined search strings. Studies were selected using the PICOC framework and evaluated through quality assessment based on methodological clarity, the explicit description of AI technologies, and their relevance to the research questions. Of 121 initially identified records, 53 eligible open-access articles were included and analyzed thematically. The findings revealed a substantial increase in publications, from one article in 2021 to 25 articles in 2025. AI chatbots were the most frequently examined technology, followed by other AI-based platforms, adaptive learning systems, intelligent tutoring systems, and machine-learning applications. AI supported mathematical problem solving by providing scaffolding, immediate feedback, strategy guidance, and systematic visualization of solution processes. It also promoted self-directed learning through personalized pathways, flexible access, adaptive feedback, learning-pace regulation, and increased confidence. Nevertheless, only a limited number of studies examined both outcomes simultaneously. Significant challenges included teachers’ insufficient digital literacy, unequal access, additional instructional workload, students’ overreliance on AI, data-privacy concerns, and algorithmic hallucinations. Future research should develop empirically validated pedagogical models, conduct meta-analyses, and establish ethical guidelines to ensure that AI functions as a safe, inclusive, and sustainable cognitive facilitator in mathematics learning.]]>
