Gestational Diabetes Mellitus (GDM) is a metabolic disorder affecting 15% of pregnancies globally, increasing maternal risks such as gestational hypertension, pre-eclampsia, caesarean delivery and Type 2 Diabetes Mellitus (T2DM). Infants born to GDM-affected mothers face higher risks of macrosomia, obesity, and metabolic complications. The standard diagnostic method, oral glucose tolerance test (OGTT), is often conducted late in pregnancy, delaying early intervention. Additionally, the COVID-19 pandemic disrupted screening processes, further complicating timely diagnoses. Given its rising pervasiveness, GDM imposes a significant economic burden, necessitating improved awareness, diagnosis, and management strategies. By utilizing clinical, demographic, and biochemical data, developments in machine learning (ML) and deep learning (DL) present potential approaches for the early prediction of GDM. However, challenges such as imbalanced datasets, model interpretability, and limited generalizability hinder their widespread clinical adoption. This review bridges traditional clinical research with AI-driven methodologies, evaluating statistical, ML, and DL models for GDM detection by exploring dataset limitations and scalability issues in current methodologies. By synthesizing medical and AI-based research, this paper provides valuable insights for healthcare professionals and researchers by giving the statistics which model performed best. This paper will help the researchers to work on the future research, and should enhance model transparency, dataset diversity, and AI integration for scalable, efficient GDM management.