Secure land ownership forms the foundation of economic development, investment confidence, social stability, and effective land governance. However, historical inconsistencies in cadastral surveys, paper-based land records, boundary re-surveys, inheritance claims, and administrative documentation continue to generate ownership disputes that delay property transactions, increase litigation, and weaken public confidence in land administration systems. As land information repositories become increasingly digitised, substantial opportunities have emerged to transform decades of archived survey records into predictive intelligence for dispute prevention rather than post-conflict resolution. This study investigates the application of machine learning techniques for detecting potential land ownership disputes through the systematic analysis of historical survey records, cadastral maps, title registrations, boundary adjustment documents, and geospatial parcel histories. The proposed analytical framework integrates natural language processing for legal document interpretation, graph-based relationship modelling for ownership lineage reconstruction, anomaly detection for identifying conflicting survey records, and supervised classification algorithms for dispute risk prediction. Feature engineering incorporates temporal ownership transitions, parcel geometry inconsistencies, surveyor records, registration frequency, and historical boundary modifications to improve predictive accuracy. Explainable artificial intelligence is further employed to enhance transparency by identifying the factors contributing to each dispute prediction, thereby supporting legal defensibility and stakeholder trust. The study demonstrates that machine learning can transform historical cadastral archives into proactive decision-support systems that strengthen land administration, reduce litigation, accelerate property transactions, and improve tenure security through evidence-based dispute prevention.