We present a data-informed approach to parametric aerodynamic shape optimization with latent diffusion. Generative models have shown success in shape design applications, where these models are often used as novel shape parameterization methods. We take a different perspective: instead of introducing new parameterization methods that substitute existing approaches, we empower existing approaches through a latent diffusion process. The construction of our method is based on the fact that virtually any shape parameterization method that maps between parameter space and shape space can be considered as a shape en-/decoder. In this work, we demonstrate our method using aerodynamic shape optimization of airfoils. We use the Hicks-Henne method (HHM) for shape parameterization, the UIUC airfoil database as training data, SU2 as the CFD solver, and Bayesian Optimization as optimizer.