Aims: We assess the sensitivity of void shapes to the nature of dark energy that was pointed out in recent studies. We investigate whether or not void shapes are useable as an observational probe in galaxy redshift surveys. We focus on the evolution of the mean void ellipticity and its underlying physical cause. Methods: We analyse the morphological properties of voids in five sets of cosmological N-body simulations, each with a different nature of dark energy. Comparing voids in the dark matter distribution to those in the halo population, we address the question of whether galaxy redshift surveys yield sufficiently accurate void morphologies. Voids are identified using the parameter free Watershed Void Finder. The effect of redshift distortions is investigated as well. Results: We confirm the statistically significant sensitivity of voids in the dark matter distribution. We identify the level of clustering as measured by sigma_8(z) as the main cause of differences in mean void shape <epsilon>. We find that in the halo and/or galaxy distribution it is practically unfeasible to distinguish at a statistically significant level between the various cosmologies due to the sparsity and spatial bias of the sample.