This study presents an AI-enhanced geodetic framework for detecting pre-seismic ionospheric disturbances using GNSS-derived Total Electron Content (TEC). The proposed approach integrates multi-source geospatial data, including GNSS observations (RINEX), global ionospheric maps (IONEX), seismic catalogs, and space weather indices (e.g., Kp, Dst, F10.7). A robust data processing pipeline is developed, including TEC extraction, noise filtering, detrending, and advanced feature engineering (e.g., ROTI, S4, TEC gradients). Machine learning techniques such as Isolation Forest, One-Class SVM, Autoencoders, and deep learning time-series models (LSTM, CNN-LSTM) are applied to identify subtle ionospheric anomalies in complex and non-stationary data. Results from case studies in seismically active regions demonstrate that significant TEC anomalies can be detected 2–5 days prior to earthquake events. To reduce false detections, the framework explicitly incorporates space weather parameters, enabling discrimination between seismic-related disturbances and solar-terrestrial effects. The proposed system highlights the potential of AI-enhanced GNSS sensing for geohazard monitoring and contributes to the development of future earthquake early-warning support systems within the broader context of Earth system observation.