Abstract
The proliferation of metaheuristic optimization algorithms raises questions about methodological novelty, validation, reproducibility, and practical applicability. This systematic review characterizes English-language journal articles introducing new metaheuristics, published from January 2020 to 4 July 2026 and retrieved through the Scopus title query TITLE (novel AND metaheuristic AND algorithm). The analysis examines publication trends, algorithmic characteristics, reported validation practices, implementation resources, and applications. Among complete publication years (2020–2025), annual counts increased overall, with fluctuations, and peaked in 2025. Human-based, bio-inspired, and swarm-intelligence approaches predominate, although their relative representation is sensitive to venue composition. Continuous, single-objective, population-based methods with uniform random initialization are prevalent; these broad descriptors do not establish similarity between search operators or underlying mechanisms. Evaluation relies heavily on conventional benchmarks, recurrent CEC suites, and a limited set of engineering problems. Statistical procedures and code availability vary across studies; their reported presence does not establish methodological rigor or computational reproducibility. Problems recorded under the domain-specific category were less frequently identified than benchmark functions; this category includes application-derived benchmarks. Analyses excluding the dominant journal and stratifying by journal quartile characterize variation associated with venue composition. The corpus combines taxonomic diversity with recurrent optimization settings and evaluation practices, highlighting opportunities for clearer methodological differentiation, transparent experimental protocols, accessible implementations, and broader application-oriented evaluation. These findings describe publications retrieved by the specified search strategy and may not generalize to newly proposed metaheuristics as a whole.