The aim of this paper is twofold: (i) to deduce the most representative C 2 N profile(s) for Dome C (DC), Antarctica, from the latest measurements, and (ii) to evaluate the performance of a widefield adaptive optics (AO) system equipping a 2-3 m telescope. Two models of the C 2 N profile, corresponding to the bimodal distribution of seeing (a poor seeing mode and a good seeing mode), are composed from both Single Star Scidar data and balloon radio soundings. The anisoplanatic error is first evaluated for a standard AO system from Monte Carlo simulations. DC is shown to outperform Mauna Kea for both seeing modes. A simple ground-layer AO (GLAO) system is then considered. This provides an anisoplanatic error of less than 150 nm over a field of 30 arcmin for the good seeing mode, corresponding to a basic performance Strehl ratio (considering also the fitting and the servo-lag errors) of more than ∼80 per cent in K and ∼50 per cent in J. The poor seeing model shows performance comparable to the Mauna Kea model. We also studied the influence of telescope elevation, showing that a telescope at 40 m would perform, in the poor seeing mode, like a telescope observing 8 m above the ground in the good seeing mode. Finally, we show that while tip-tilt-only correction permits high levels of correction in the good seeing mode at 40 m, it is not as efficient as the GLAO system, even at an altitude of 8 m, and it is not sufficient for high levels of correction for poor seeing, even at a height of 40 m.