Because of the widespread use of additives, a revised dynamic modulus (|E*|) predictive approach is needed for the mechanistic–empirical analysis of asphalt pavement designs. The importance of predicting |E*| relates to its use as a sensitive input in the AASHTO Pavement ME Design software. The objective of this study is to evaluate two |E*| predictive models currently used for asphalt mixtures that include non-conventional additives, and to provide recommendations for enhancing prediction capability of the models when these types of materials are used. The focus of the paper is on finding ways to efficiently and accurately generate |E*| inputs for the software when actual pavement response is not measured. Impacts of mixture type, temperature, and binder grade on |E*| predictions are explored. The current software uses three levels of inputs, two of which require key material characteristics to predict |E*| values rather than employing laboratory-measured |E*|. This study shows that for non-conventional mixtures, 53 % of |E*| values predicted by the 2006 Witczak predictive equation (WPE) are greater than values predicted by the 1999 WPE. This trend of over-prediction is observed at both low and high test temperatures. However, the opposite effect exists between the two WPEs at intermediate test temperatures. This research shows that adjusting the original binder type of non-conventional mixtures provides a significant improvement in predictive accuracy for |E*|. The findings of this study provide suggestions for alterations to binders, to adjust the WPE model’s terms for better predictive accuracy, and supplies recommendations to practitioners who use the AASHTOWare Pavement ME Design software; however, it is still evident that the laboratory measurement of |E*| values is still the preferred approach for generating inputs for the software.