To systematically review the current landscape of Artificial Intelligence (AI) and Machine Learning (ML) applications in epilepsy care, with a focus on diagnostic, predictive, and therapeutic innovations, particularly in pediatric populations.

The integration of AI/ML into epilepsy management is accelerating, driven by the proliferation of digitized medical data (EEG, MRI) and computational advancements. These technologies offer transformative potential in automating diagnostics, forecasting seizures, and personalizing treatment, yet their clinical adoption remains uneven.

A comprehensive literature search was conducted across PubMed, Embase, IEEE Xplore, and Scopus for studies published between January 2015 and September 2025. Inclusion criteria encompassed original research applying AI/ML to epilepsy diagnosis, seizure detection/prediction, or treatment support. Data were extracted on algorithm type, input modality, performance metrics, and validation rigor.

From 1,243 screened articles, 78 met inclusion criteria. Key findings include:

• Diagnosis: Deep learning models (CNNs, DeepSOZ) achieved AUROCs of 0.887–0.996 in automated EEG interpretation and lesion localization.
• Seizure Detection/Forecasting: Multimodal sensor systems (EEG, accelerometry, electrodermal activity) demonstrated sensitivities of 92–100% and low false alarm rates (0.2–1/day). Forecasting models identified pre-ictal states with sensitivities up to 87.8%.
• Therapeutic Management: ML predicted ASM response and surgical outcomes with accuracies exceeding 89%. Neurostimulation devices like RNS utilized AI for real-time seizure modulation.
• Methodological Rigor: Only 15% of studies employed prospective or pseudoprospective validation. Personalized models and standardized data frameworks (e.g.,the Observational Medical Outcomes Partnership Common Data Model (OMOP CDM)) were critical for scalability and reproducibility.

AI/ML technologies are reshaping epilepsy care across diagnostic, predictive, and therapeutic domains. While technical performance is promising, clinical translation requires robust validation, personalized modeling, and integration into existing workflows. Future research should emphasize real-world trials, regulatory pathways, and clinician-AI collaboration.