Traditional static tests of reading and reading-related skills offer some ability to predict future reading performance, though such screeners may misclassify children with or at risk of reading disorder (RD). Dynamic assessment (DA) is an alternative approach that measures learning potential and may be less dependent on learning background. A systematic review was carried out to examine the ability of DA to classify children with or at risk of RD. A database search yielded 14 eligible articles assessing DA of decoding, phonological awareness (PA), and working memory (WM). Results suggest that DA explains unique variance in the prediction of later RD status, and although models with a single dynamic predictor sometimes achieved good classification accuracy, this was enhanced somewhat by the addition of static predictors. Higher classification accuracy was found for DA targeting constructs more proximal to reading, particularly decoding, but the predictive power of DA of decoding and PA appeared to wane with increasing age as static measures explained more variance in outcomes. Some evidence emerged that DA provides benefits over static tests for the prediction of RD in bilingual students, though no studies examined outcomes by administration format or orthographic depth. Limitations and suggestions for future work are discussed.