Three prototype applications in three different media were constructed during this research. The three different media were desktop computers, interactive TV and mobile devices. These three applications incorporated methodologies based on user modelling and adaptivity. User models were constructed in order to help users in remote shopping applications in two separate ways. Firstly, by automatically recommending products close to their interests and secondly by helping them adaptively while they used the application. Explicit information, observation of user behaviour and dynamic stereotypes were incorporated into the construction of these user models. The incorporation and combination of these technologies into three different media constitutes a very novel approach on the topic of personalization in remote shopping applications. The three prototypes also incorporated techniques of adaptivity for visualizing the recommendations based on interest and needs and for personalizing the user interaction with the application. The methodology used was a combination of adaptive hypermedia and dynamic user interface and proved to be a very important aspect in the personalization and familiarity of the applications. Furthermore, these methodologies were shown very effectively for the purpose of dynamically assisting users. A major research field of this research is the field of machine learning. More specifically, recommendation applications that incorporate machine learning algorithms in their reasoning process. These applications recommend products or processes to users according to their interests or similar users’ interests. The field of recommendation applications combined with machine learning algorithms has been popular research topic in the recent years and significant results have been achieved.