The accurate representation of bidirectional ammonia (NH₃) biosphere–atmosphere exchange is an important part of modern air quality models. However, the cuticular (or external leaf surface) pathway, as well as other non-stomatal ecosystem surfaces, still pose a major challenge to translating our knowledge into models. Dynamic mechanistic models including complex leaf surface chemistry have been able to accurately reproduce measured bidirectional fluxes in the past, but their computational expense and challenging implementation into existing air quality models call for steady-state simplifications. Here we qualitatively compare two semi-empirical state-of-the-art parameterizations of a unidirectional non-stomatal resistance (R_w) model after Massad et al. (2010), and a quasi-bidirectional non-stomatal compensation-point (χ_w) model after Wichink Kruit et al. (2010), with NH₃ flux measurements from five European sites. In addition, we tested the feasibility of using backward-looking moving averages of air NH₃ concentrations as a proxy for prior NH₃ uptake and as a driver of an alternative parameterization of non-stomatal emission potentials (Γ_w) for bidirectional non-stomatal exchange models. Results indicate that the R_w-only model has a tendency to underestimate fluxes, while the χ_w model mainly overestimates fluxes, although systematic underestimations can occur under certain conditions, depending on temperature and ambient NH₃ concentrations at the site. The proposed Γ_w parameterization revealed a clear functional relationship between backward-looking moving averages of air NH₃ concentrations and non-stomatal emission potentials, but further reduction of uncertainty is needed for it to be useful across different sites. As an interim solution for improving flux predictions, we recommend reducing the minimum allowed R_w and the temperature response parameter in the unidirectional model and revisiting the temperature-dependent Γ_w parameterization of the bidirectional model.