The intersection of metaheuristic optimization algorithms and Artificial Intelligence (AI) has emerged as a transformative research frontier, yielding significant advances in the automated design and tuning of Machine Learning (ML) models. This paper presents a comprehensive systematic review, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, examining 347 peer-reviewed studies published between 2015 and 2025 across five major scholarly databases: Scopus, Web of Science (WoS), IEEE Xplore, ACM Digital Library, and arXiv. The review investigates three critical domains of AI optimization where metaheuristic algorithms have demonstrated exceptional efficacy: 1) Neural Architecture Search (NAS), encompassing convolutional, recurrent, and transformer architecture design, 2) Hyperparameter Optimization (HPO), covering learning rate tuning, batch size selection, regularization parameter calibration, and optimizer configuration, and 3) intelligent feature engineering, including wrapper-based feature selection, feature construction, and dimensionality reduction. Our analysis reveals that evolutionary algorithms (Genetic Algorithms (GAs), Differential Evolution (DE)) and swarm intelligence methods (Particle Swarm Optimization (PSO), Grey Wolf Optimizer (GWO), Whale Optimization Algorithm (WOA)) consistently outperform traditional grid search and random search methods, achieving average accuracy improvements of 2.3%–5.8% while reducing computational cost by 40–75%. Furthermore, hybrid metaheuristic–AI approaches demonstrate synergistic performance gains exceeding those of standalone methods. The review also provides bibliometric analysis, identifies key research trends, highlights methodological challenges—including computational overhead, scalability limitations, and reproducibility concerns—and proposes eight future research directions spanning federated optimization, quantum-inspired metaheuristics, and Large Language Model (LLM) architecture search. This work serves as a comprehensive reference for researchers and practitioners seeking to leverage metaheuristic intelligence for automated AI model optimization.