<p>Dowel-type fasteners are widely used to connect cross-laminated timber (CLT) members, and the embedment strength of CLT therefore plays a critical role in governing the performance of these connections. Reliable design requires a sound understanding of embedment behavior associated with dowel-type fasteners and the ability to characterize it accurately. This study reviewed current methods for testing CLT embedment strength, discussed the failure modes of CLT specimens under embedment loads, and examined the factors influencing dowel embedment strength. A database of 920 half-hole embedment tests was used to develop machine learning (ML) models for the reliable prediction of CLT embedment strength. The predictive performance of the ML algorithms was compared with that of theoretical models. Based on a trained extreme gradient boosting (XGBoost) model, Shapley additive explanations (SHAP) analysis was conducted to evaluate the influence of input features on CLT embedment strength from both global and local perspectives. The modeling results indicated that specimen density, parallel layer ratio, and dowel diameter had the greatest influence on embedment strength, with density being the predominant factor. The XGBoost model was retrained using these three features as inputs, with predictions for generated virtual specimens showing good agreement with those obtained using six input features.</p>