This paper examines how AI-driven gamification can improve student motivation and engagement in STEM education. It reviews major theories, synthesizes current empirical research, and evaluates how adaptive algorithms, learning analytics, and conversational agents enhance traditional gamified learning. The study highlights where AI-based gamification is effective, where evidence is still limited, and identifies the methodological gaps that remain in the field. Key findings indicate that AI personalization can enhance intrinsic motivation, foster sustained behavioral and cognitive engagement, and support student progress through adaptive feedback. The review highlights several challenges, including the use of short-term studies, overreliance on self-report data, equity concerns, privacy issues, and inconsistent implementation across various learning environments. The review provides clear implications for educators, policymakers, and developers who aim to utilize AI-enhanced gamification responsibly. It emphasizes the need for long-term research, better measurement practices, and equity-centered design to ensure that AI benefits all learners in STEM.