The reliability of the results obtained by applying the Generalized Inversion Technique (GIT) to site response estimates is assessed by means of numerical simulations. This study was developed in the framework of the UMBRIA-MARCHE microzonation project. The analysis with synthetic data demonstrates that the generalized inversion technique is capable of revealing the site effects. Furthermore, this approach shows that a priori information relevant to the Q quality factor is needed if a source estimate is required. We also analyze data recorded by five seismological stations installed in Fabriano (central Italy) with a minimum interstation distance of 100 m. The results computed with the generalized inversion technique are compared with those obtained by the reference site method and the receiver function technique. Whereas the first two methods provide highly consistent site-response estimates, the receiver function technique provides lower levels amplification. Furthermore, the time duration used in the analysis seems to be a critical parameter for the receiver function technique alone.