Lake water temperature is an important indicator to characterize the ecological environment of lakes, affecting the physical, chemical, and biological processes occurring in lakes. Investigating the evolution of lake water temperature is of great significance for the management and improvement of lake ecological environment. A Long Short-Term Memory (LSTM) neural network was used to construct a reconstruction model for the water temperature of Dongting Lake. The performance of variables including air temperature (T), day of year (DOY), and discharge (Q) was evaluated in the reconstruction of lake water temperature. Using the LSTM model, the long-term water temperature series of the Dongting Lake Basin was reconstructed, and the characteristics of water temperature changes from 1960 to 2020 were analyzed. The study also explored the impacts of different meteorological datasets as supplementary inputs on the model's performance in simulating water temperature. The results showed that: (1) The LSTM neural network-based water temperature reconstruction model demonstrated strong performance in simulating water temperature. Incorporating both DOY and Q as input variables enhanced the model’s accuracy, with DOY contributing more significantly to the improvement than Q. (2) From 1960 to 2020, the annual average water temperature in the Dongting Lake Basin showed a warming trend, with an average warming rate of 0.15°C/10a. However, there were spatial differences in the warming rates, with the lake area warming faster than the main tributaries. On a seasonal scale, there were significant seasonal variations in the warming rates, with spring being the season with the fastest warming, at an average rate of 0.29°C/10a, while summer had a more moderate warming rate of 0.04°C/10a. (3) The annual average water temperature and air temperature in the Dongting Lake Basin and its main tributaries exhibited similar abrupt change characteristics, both undergoing abrupt changes around 1996. The abrupt change in air temperature was a significant factor driving the abrupt change in water temperature. (4) Meteorological factors and lagged information, as supplementary inputs to the water temperature reconstruction model, contributed to varying degrees of improvement in model performance. However, the extent of performance enhancement showed regional differences, which may have been due to variations in the driving mechanisms of water temperature changes and the relationship between water and air temperature responses in different regions. The findings of this study are significant for protecting the water environment of the Dongting Lake Basin and offer new insights into the study of water temperature evolution in lakes with limited observational data.