Aural interactive web sites are mainly (and in same cases exclusively) based upon "listening to", instead of "looking at". They are needed when users have visual impairments, or not much attention can be paid to a screen (possibly also a tiny one). Web sites, today, and even most mobile applications, use, above all, the "visual channel" to communicate content, navigation/interaction capabilities and "interface messages" (such as orientation, application structure, page structure, priorities of elements on the page, ). However, requirements for aural web sites, despite their relevance (e.g. W3C-WAI initiatives), are poorly addressed by current accessibility guidelines. Most of the focus, in fact, seem to lay upon basic accessibility issues (such as the fact that each image should have a semantically equivalent text), or about technicalities (such as how to avoid the use of tables for the pages layout). This paper discusses salient requirements for aural web sites and their implications upon both the user experience and the web sites evolution. The work is based upon basic (linguistic) research and upon real-life experience, in which web sites for visuallyimpaired users and mobile scenarios have been designed and also experimented.