Improvement of wireless power transfer (WPT) systems is necessary to tackle issues of power transfer efficiency, high costs due to sensor and communication requirements between the transmitter (Tx) and receiver (Rx), and maintenance problems. Analytical techniques and hardware-based synchronization research for Rx-sensorless WPT may not always have been available or accurate. To address these limitations, researchers have recently employed machine learning (ML) to improve efficiency and accuracy. The objective of this work was to replace Tx–Rx communication with ML, utilizing Tx-side parameters to predict the load and coupling coefficients on an LC–LC tuned WPT system. Based on current and voltage features collected on the Tx-side for various load and coupling coefficient values, we developed two models for each load and coupling prediction. This study demonstrated that the extra trees regressor effectively predicted the characteristics of LC–LC tuned WPT systems, with coefficients of determination of 0.967 and 0.996 for load and coupling, respectively. Additionally, the mean absolute percentage errors were 0.11% and 0.017%.