Two probabilistic seismic hazard (PSH) maps in terms of macroseismic intensity characterized by an exceedance probability of 10% for exposure time of 50 years are presented and compared. The first map adopts the standard Cornell–McGuire approach and follows the computational scheme developed for the reference Italian peak ground acceleration (PGA) hazard map (MPS04), while the second one is derived through an alternative methodology (referred to here as the site approach) that is based on statistical analysis of the site seismic history (i.e., macroseismic intensities documented for past earthquakes). Because the two procedures make a different use of available information, this comparison provides a new insight about the sensitivity of PSH estimates for the different possible methodological choices. In particular, it is shown that, though basic differences exist between the two adopted methodologies, relevant results appear consistent over most of Italy. However, at a significant number of investigated localities (Italian municipalities), PSH estimates provided by the site approach are larger than those derived from the standard technique. Thus, a detailed analysis has been carried out to evaluate the role played by different choices of computational models and input data. Among these, the use/nonuse of seismogenic zoning seems to act as the key element in determining the pattern of differences observed between the two PSH estimates.