In this paper, we develop a predictive model for ionospheric Total Electron Content (TEC) using a straightforward multivariate Long Short-Term Memory (LSTM) machine learning algorithm. Time series data from a single GNSS station over eight years was used for training (90%). The model was trained on data from 2012–2019, covering both maximum solar activity from 2012–2016 and minimum solar activity from 2017 - 2019. The training process included solar and geomagnetic indices F10.7, Kp, and Dstas model inputs. The Ionosphere Weather Index (WTEC Index), an indicator of ionospheric changes due to geomagnetic disturbances, was also incorporated. The LSTM model was developed to capture seasonal variations on a monthly basis. The RMSE was approximately 4 TECU, calculated using (10%) of the test data year 2020, with the maximum observed TEC values ranging from 50 to 60 TECU. However, testing and validating the model using the 2022 observation data showed a maximum RMSE below 11 TECU (with maximum TEC observation data ranging from 100 to 110 TECU). The model’s RMSE is reasonably accurate and can be adjusted with a 10% correction factor. The model’s trend variations generally match. The data is observational, with a correlation coefficient of up to 0.8. A comparison of the model predictions for the IRI model revealed that it underestimates both the LSTM model and observation data. The evaluation demonstrates that the LSTM model effectively responds to geomagnetic storms, whether moderate or intense.