Any analysis of paths on the Internet has to include a detailed description of the pages visited. We propose a methodology for analysing hypertext documents, based on an identification of structural features (structure of links, description of elements of the page), presentation (characters used, screen background, etc.) and content (personal pronouns, text content, etc.). By applying this method to corpuses of sites we are able to identify the specific characteristics of personal sites compared to commercial sites and thus further our understanding of the articulation between the non-commercial and commercial spheres. If we then apply the method to a corpus of personal pages visited, we can differentiate between types of documents and understand the link between page characteristics and the way in which visits are organized.