The study developed and validated a machine learning model to predict high-risk distant metastatic recurrence in patients with differentiated thyroid carcinoma (DTC). They analyzed 1245 patients, of which 126 (10.1%) developed a recurrence during a median follow-up of 72 months. Using LASSO regression, they identified eight predictors: age, tumor size, extrathyroidal extension, lymph node metastasis, BRAF V600E mutation, postoperative stimulated thyroglobulin (sTg), radioactive iodine dose, and TNM stage. Of the six algorithms (including Random Forest, XGBoost, and Logistic Regression), the XGBoost model performed best with an AUC of 0.88 (95% CI, 0.83–0.93) on the validation set. Patients were divided into low-risk (recurrence 1.7%), intermediate-risk (14.4%) and high-risk (64.1%) groups, with significantly different metastasis-free survival (p < 0.001). The model demonstrated good calibration and higher clinical utility than the TNM system. The model can help identify patients for more aggressive treatment and more intensive follow-up.