<p>Recently, bio-inspired metaheuristic algorithms have been widespread in various application areas due to their straightforward implementation and ability to handle and solve complex problems. Despite the widespread adoption of metaheuristic algorithms, there is a lack of adaptation of the artificial rabbit optimization (ARO) algorithm to solve classification tasks. Specifically, the ARO algorithm has primarily been used for hyperparameter tuning machine learning algorithms or to select the best feature subset within a wrapper feature selection without cooperating in the classification process. This paper has introduced a new direct adaptation of the ARO algorithm for classification tasks. This adaptation was constructed by directly using the ARO algorithm to find the optimal centroids that could solve the classification task. Therefore, the ARO algorithm was shifted directly to the classification task by identifying the optimal centroid for each class label and minimizing the number of misclassified instances in the training data. The proposed direct adaptation of the ARO algorithm for classification tasks was tested and evaluated using eleven benchmark datasets from various domains. We conducted in-depth investigations of the other seven bio-inspired optimization algorithms alongside the ARO algorithm. The accomplished results demonstrated the robustness and effectiveness of the ARO algorithm for solving the classification problem compared to other bio-inspired optimization algorithms. Consequently, by extending ARO to focus on optimal centroid identification, our approach has enhanced the performance of classification tasks, thereby improving the effectiveness of decision-making models.</p>