We use the most recent Type Ia supernova data in order to study the dark energy–dark matter (XCDM) unification approach in the context of the generalized Chaplygin gas (GCG) model. Rather surprisingly, we find that data allow models with α > 1. We have studied how the GCG adjusts flat and non-flat models, and our results show that GCG is consistent with a flat case up to 68 per cent confidence level. Actually this holds even if one relaxes the flat prior assumption. We have also analysed what one should expect from a future experiment such as the Supernova/Acceleration Probe (SNAP). We find that there is a degeneracy between the GCG model and an XCDM model with a phantom-like dark energy component.