One of the primary challenges in the development of high-speed systems is accurate and efficient prediction of the aerodynamic heating, particularly for light-weight systems where the aerothermodynamic loads are strongly coupled with the structural response. This study builds upon previous work focused on using CFD surrogates for prediction of aerothermodynamic loads within a tightly integrated fluid-thermal-structural analysis framework. A novel approach is developed that corrects heat flux predictions from a pointwise dependency on surface temperature in order to account for surface temperature gradients. This enables efficient construction of a CFD surrogate for aerodynamic heating without a priori assumptions on the surface temperature profile; while also accurately maintaining the critical feedback behavior of the surface temperature profile for a coupled fluid-thermal-structural analysis. The method is compared, in terms of accuracy and efficiency, with a hierarchy of approaches for aerodynamic heating prediction. Comparison cases include simple flat plate heating, as well as shock impingements in two dimensional and three dimensional flows. Typically the approach yields less than 5% error relative to full-order CFD predictions, and clearly outperforms all other simple prediction methods in terms of accuracy. Furthermore, it maintains excellent computational efficiency, enabling long time record fluid-thermal-structural analysis in complex, three-dimensional flow environments.