Objective: Syncope is defined as a sudden loss of consciousness due to cerebral hypoperfusion, with vasovagal syncope (VVS) being the most common form. The Head-Upright Tilt Test (HUTT) is the primary diagnostic tool but is time-consuming and has a suboptimal diagnostic yield. Machine Learning (ML) may improve early syncope prediction, thereby increasing diagnostic efficiency and reducing the burden on patients and healthcare professionals. Methods: We searched PubMed for studies using ML on HUTT data for syncope testing. Extracted data included ML models, input features, performance metrics, preprocessing, and evaluation methods. Study quality was assessed using the STAR-ML checklist. Results: Thirteen studies were included. Commonly used ML algorithms were support vector machines (SVM), neural networks, decision trees, k-nearest neighbor, and logistic regression. Features were derived from Electrocardiogram (ECG), continuous blood pressure (CBP), and transthoracic impedance (TIM). The highest-performing model used an SVM with features from ECG, CBP, and TIM. Conclusions: ML integrated with HUTT signal analysis shows promise for improving diagnostic accuracy and efficiency. SVM models using multimodal features were particularly effective. Significance: This review supports further development of ML-based tools to enhance diagnostic workflows in syncope care, especially for early VVS prediction.