A remote descriptive analysis panel was formed and trained over the Internet. This paper outlines the steps involved in this method: (1) creation of Web sensory sessions; (2) recruitment and selection of assessors; (3) generation of descriptors; (4) training using generated descriptors; and (5) check-up sessions. A reference panel was formed and trained conventionally: it aimed, on the one hand, at providing the necessary reference points for training the Internet panel and on the other hand, at validating the proposed method by the comparison of the results assessed on the same product by both panels. We applied our methodology to a dry sausage. Sixteen assessors carried out the complete training on 10 descriptors during 10 sessions. The results of the analysis of variance with (panel-product) interactions considering each attribute separately showed that product classification is the same for the two panels on the basis of seven descriptors out of 10. The results concerned with the performance of individual assessors exhibit that training duration seemed to be the most important feature in determining the level of performance of assessors remotely trained over the Internet.