Prediction models are of a great assistance for predicting the development of a disease, detecting or screening undiagnosed patients, predicting the effectiveness of a treatment and helping toward better decision-making. Recently, three predictive scores in the field of autoimmune thyroid disease (AITD) have been introduced: The Thyroid Hormones Event Amsterdam (THEA) score: a predictive score of the development of overt AITD, the Graves’ Events After Therapy (GREAT) score: a prediction score for the risk of recurrence after antithyroid drugs withdrawal and the Prediction Graves’ Orbitopathy (PREDIGO) score: a prediction score for the development of Graves’ orbitopathy in newly diagnosed patients with Graves’ hyperthyroidism. Their construction, clinical applicability, the possible preventative measurements which can be taken to diminish the risks and the potential future developments which can improve the accuracy of the predictive scores are discussed in this review.