This paper presents an artificial intelligence-based architecture to be used in various tourism services, namely the user modeling and recommender system components. Through the use of several machine learning techniques, such as linear models, neural networks, classification and even text mining, a hybrid and complete approach at understanding tourism application users is attained. To assess the advantages of the model, a specific prototype was created, including a complex user modeling process and an advanced and innovative recommender system. The system was tested in the scope of Oporto city, in Portugal.