<p>The Covariance Matrix Adaptation Evolution Strategy (CMA-ES) is a benchmark in evolutionary computation, widely used for continuous black-box optimization, with strong empirical performance in many benchmark settings. Despite its impact, its practical performance remains highly dependent on specific structural configurations and parameter settings. To address these challenges, a vast ecosystem of modifications has been developed. Following a systematic review of over 170 peer-reviewed publications—selected from high-impact databases based on their citations and relevance—this survey introduces a novel hierarchical taxonomy of CMA-ES modifications and hybridizations. Beyond a descriptive review, this work provides a critical synthesis of the algorithm’s evolution, offering a structured framework to analyze performance trade-offs and structural synergies. The analysis reveals that while recent hybrids have been reported to improve convergence speed and/or exploration in specific regimes and benchmarks, significant gaps remain in scalability and automated parameter tuning. Across the surveyed literature, a consistent trend is that mechanism-preserving refinements tend to be the most transferable across benchmarks, whereas regime-specific extensions (e.g. surrogate assistance, uncertainty handling, and constraint management) yield the largest gains mainly under expensive, noisy, or feasibility-limited settings. This approach maps current trends and supports the development of future variants, highlighting scalable covariance adaptation and automated parameter tuning, while providing a practitioner-oriented decision guide for navigating the CMA-ES literature.</p>