<p>Fiber-reinforced polymer (FRP) composites play a pivotal role in aerospace, automotive manufacturing, and medical industries due to their exceptional specific strength and lightweight properties. However, traditional manufacturing optimization faces challenges such as inefficiency and high costs. Machine learning (ML), with its data-driven predictive capabilities, is emerging as a powerful tool in this field. This paper systematically reviews the latest applications of ML in the manufacturing processes of fiber-reinforced polymer composites, covering its role in critical thermoset manufacturing processes (e.g., autoclave curing, liquid composite molding, filament winding) and thermoplastic manufacturing processes (e.g., additive manufacturing, injection molding). The analysis indicates that, despite significant progress, the application of ML in composite manufacturing continues to face fundamental challenges, including the complexity of integrating ML models into real-time control loops, the scarcity of high-quality industrial datasets, and insufficient model generalization across different material systems. Therefore, future directions should focus on developing hybrid models with physics-informed fusion, establishing cross-industry benchmark datasets, and advancing edge computing solutions for adaptive manufacturing. By addressing information fragmentation and uncovering scenario-based insights, this review aims to guide the field toward a more systematic and highly reliable development path.</p> Graphical Abstract <p></p>