<p>Groundwater, the primary global source of liquid freshwater, faces a crisis due to widespread over-drafting, leading to significant declines in groundwater levels. This study predicts and forecasts monthly groundwater levels (GWLs) at three observation wells, Ramachandrapuram, Palakollu, and Jangareddigudem, in the Lower Godavari River Basin, India. In this study, the advanced Univariate Artificial Intelligence (AI) techniques, including eXtreme gradient boosting (XGBoost), light gradient-boosting machine (LightGBM), and classification and regression trees (CART) were employed for this purpose. The methodology involves extracting time-dependent features from groundwater data through time series analysis. Models were developed using data from a training period (Jan 1998–Jun 2008) and evaluated during a testing period (Jul 2008–Dec 2012). XGBoost, LightGBM, and CART were utilised to predict monthly GWLs, and these models were subsequently employed to forecast GWLs from Jan 2013 to Dec 2018. Among the models tested, XGBoost demonstrates superior performance, outperforming LightGBM and CART with R² values of 0.91, 0.89, and 0.85 for Jangareddigudem, Ramachandrapuram, and Palakollu, respectively, during the testing period. It highlights XGBoost’s efficacy in capturing complex relationships within groundwater data, thus enhancing predictive accuracy. The innovative data-driven modelling approach proposed in this study facilitates accurate GWL estimation and supports sustainable policy decisions, particularly in regions with limited data availability. This research contributes to the broader goal of effective groundwater management and conservation by leveraging advanced AI techniques.</p>