Site characterization is an important task in geotechnical engineering practice. The ultimate goal in site characterization is to be able to estimate in situ soil properties at any half-space point at a site based on limited tests. This estimate may be a point estimate or in terms of some statistical parameters. Geostatistical and random field methods have been applied with various degrees of success. This paper presents a new approach, based on artificial neural network, for site characterization. Emphasis is placed on application of generalized regression neural networks for site characterization. The results show that neural network approach has a potential to be a practical tool for site characterization.