Context: Artificial intelligence has established itself as an emerging resource to transform teaching through personalization, tutoring, learning analytics, automated assessment, and generative tools. However, its incorporation requires pedagogical, ethical, and institutional foundations. Objective: To analyze the findings, tensions, and directions of educational research on artificial intelligence relevant to teaching innovation. Methodology: A systematized literature review with a qualitative narrative synthesis was conducted, following the PRISMA 2020 guidelines. The search considered publications between 2016 and 2026 related to teaching, learning, assessment, faculty development, literacy, and educational governance. From 43 identified records, 27 documents were included in the qualitative synthesis. Results: The findings demonstrate that personalization and tutoring can favor learning when they respond to explicit didactic objectives. Likewise, analytics facilitates the identification of patterns, while generative artificial intelligence demands a redesign of assessment strategies. Digital literacy, teacher training, and ethical governance emerge as fundamental conditions. Conclusion: Educational innovation through artificial intelligence depends on its pedagogical integration, human oversight, empirical evaluation, and institutional frameworks that guarantee equity, privacy, transparency, and accountability.