The detection of epileptic seizures is a crucial aspect of epilepsy care, requiring precision and reliability for effective diagnosis and treatment. Seizure detection plays a critical role in healthcare informatics, aiding in the timely diagnosis and management of epilepsy. The use of computational intelligence and optimization techniques has shown significant promise in improving the performance of automated seizure detection systems. This research work proposes a novel hybrid approach that combines Ant Colony Optimization (ACO) for feature selection with Gray Wolf Optimization (GWO) to optimize the hyperparameters of a Random Forest (RF) classifier. In this patient-specific seizure detection, ACO effectively reduces the feature set, improving computational efficiency, while GWO ensures optimal RF performance. The method is evaluated on the Children’s Hospital Boston–Massachusetts Institute of Technology (CHB–MIT) and Seina datasets, which include multichannel EEG data from epileptic patients. Performance metrics such as accuracy, sensitivity, and specificity are employed to evaluate the effectiveness of the seizure detection system. Results The proposed ACO-GWO-RF pipeline demonstrated excellent performance on the CHB-MIT dataset, with a mean accuracy of 96.70%, mean sensitivity of 92.66%, and mean specificity of 99.24%, outperforming existing approaches. The mean values of accuracy, sensitivity, and specificity obtained using the Seina dataset are 93.01%, 89.82%, and 96.26%, respectively. These improvements highlight the robustness of the hybrid metaheuristic method in handling complex EEG data. The hybrid metaheuristic approach effectively optimizes the processing and classification of EEG data for seizure detection. Its strong performance across datasets suggests potential for integration into interactive health applications. Furthermore, its patient-specific adaptability makes it a promising tool for personalized epilepsy diagnos