Metaheuristic optimization algorithms are widely applied in complex optimization problems due to their flexibility and ability to provide almost optimal solutions in large search spaces. One of the key challenges in these algorithms is determining an appropriate stopping criterion. Too early termination may lead to suboptimal results, while excessive execution increases computational overhead without significant improvement. This paper presents a PRISMA-based literature review of stopping criteria used in metaheuristic optimization algorithms. Following a systematic screening process, 54 primary studies were included in the final qualitative synthesis. Based on the analysis of the selected studies, we propose a taxonomy of stopping criteria based on the aspect of the metaheuristic search process used to trigger algorithm termination, covering resource-based, objective-progress-based, population-state-based, convergence-indicator-based, statistical, algorithm-specific, and hybrid approaches. The review further discusses their applicability across different metaheuristic families, identifies current research trends, and also discusses remaining research gaps.