Résumé : We aim to explore two exozodiacal dust production mechanisms, first re-investigating the Poynting-Robertson drag pile-up scenario, and then elaborating on the less explored, but promising exocometary dust delivery scenario. We developped a new versatile, numerical model that calculates the dust dynamics, with non orbit-averaged equations for the grains close to the star. The model includes dust sublimation and incorporates a radiative transfer code for direct comparison to the observations. We consider in this study four stellar types, three dust compositions, and we assume a parent belt at 50 au. We find that, in the case of the Poynting-Robertson drag pile-up scenario, it is impossible to produce long-lived submicron-sized grains close to the star. The inward drifting grains fill in the region between the parent belt and the sublimation distance, producing an unrealistically strong mid-infrared excess compared to the near-infrared excess. The dust pile-up at the sublimation radius is by far insufficient to boost the near-IR flux of the exozodi to the point where it dominates over the mid-infrared excess. In the case of the exocometary dust delivery scenario, we find that a narrow ring can form close to the sublimation zone, populated with large grains several tens to several hundred of micrometers in radius. Although not perfect, this scenario provides a better match to the observations, especially if the grains are carbon-rich. We also find that the required number of active exocomets to sustain the observed dust level is reasonable. We conclude that the hot exozodiacal dust detected by near-infrared interferometry is unlikely to result from inwards grains migration by Poynting-Robertson drag from a distant parent belt, but could instead have an exocometary origin. [Abridged]