Evapotranspiration (ETo) plays a crucial role in managing water resources and agricultural water consumption. It is also commonly used to quantify the total amount of water lost through a number of important processes that occur among the land and atmosphere. In this research, four deep learning algorithms—CNN, DNN, BiLSTM, and GRU—were applied to predict evapotranspiration based on 14 years of daily data from Victoria, a state in southeastern Australia. The data sample was split into two periods: nine years (2010–2019) for training and four years (2020–2023) for testing. Deep learning algorithms have good performance for predicting evapotranspiration. The results showed that the GRU and DNN models were slightly better than the other two models. In the testing phases, the GRU models found R-Square, RSME, MSE, and MAE values, 0.989, 0.1794, 0.0322, and 0.1417, respectively, while the DNN models performed 0.980, 0.185, 0.0345, and 0.1507 value of R-Square, RMSE, MSE, and MAE, respectively, which indicated the GRU model perform better than other models. The CNN model achieved an R² of 0.958, with an RMSE of 0.364 and an MSE of 0.1330, indicating less precise estimations. Similarly, the BiLSTM model performed better than CNN but still lagged behind GRU and DNN, with an R² of 0.969 and an MSE of 0.0988. Moreover, deep learning models perform well, the GRU model has comparatively excellent performance than other DL models. It has been suggested that the most accurate model to improve future studies on evapotranspiration estimations is the GRU model, which could improve irrigation efficiency and boost crop productivity.