Modelling techniques for the propagation of light pollution in the atmosphere allow the computation of maps of artificial night sky brightness in any direction of the sky, involving a large number of details from satellite data. Cinzano et al. (2001a) introduced a method of mapping naked eye star visibility at the zenith from large areas based on satellite radiance measurements and Garstang models of the propagation of light pollution. It takes into account the altitude of each land area from digital elevation data, natural sky brightness in the chosen sky direction based on the Garstang approach, eye capability after Garstang and Schaefer, and atmospheric extinction in the visual photometric band. Here we discuss how to use these methods to obtain maps of the average number of visible stars when looking at the night sky hemisphere, finally answering, site by site, the question of how many stars are visible in the sky. This is not trivial, as the number of stars visible depends on the limiting magnitude in each direction in the sky, and this depends on sky brightness in that direction, atmospheric extinction at that zenith distance and the observer's visual acuity and experience. We present, as an example, a map of the number of visible stars in Italy to an average observer on clear nights with a resolution of approximately 1 km.