Abstract In rock mechanics, the dilatancy point is always occurring before rock failure during loading process. Water content plays a significant role in the rock physiomechanical properties, which also impact the rock dilatancy point under loading process. This dilatancy point significantly plays a warning role in the rock engineering structures stability. Therefore, it is essential to predict the rock dilatancy point under different water contents to get an early warning for effective monitoring of engineering projects. This study investigates the water contents effects on sandstone dilatancy point under loading in the presence of infrared radiation (IR). Furthermore, this IR was used for the first time as an input parameter for different artificial intelligence (AI) techniques to predict the dilatancy point in the stress-strain curve. The experimental findings show that the stress range in stress-strain curve stages (crack closure and unstable crack propagation) increases with water content. However, this range for deformation and stable crack propagation stages decreases with water content. The dilatancy stress, crack initiation stress, and elastic modulus are negatively linearly correlated, while peak stress and stress level are negatively quadraticaly correlated with a high (R2). The absolute strain energy rate, which gives a sudden increase at the point of dilatancy, is used as the dilatancy point index. The stress level is 0.86 σmax at the dilatancy point for dry rock and decreases with water content. This index is predicted from IR data using three computing techniques: artificial neural network (ANN), random forest regression (RFR), and k-nearest neighbor (KNN). The performance of all techniques was evaluated using R2 and root-means-square error (RMSE). The results of the predicted models show satisfactory performances for all, but KNN is remarkable. The research findings will be helpful and provide guidelines about underground engineering project stability evaluation in water environments.