Social media platforms allow users to socialize with their peers by sharing information in various forms, including photos and tags. This data, extracted from these online networks, paves the way for innovative research that offers the novel possibility of proposing personalized recommendations, for potential travel destinations based on user-generated traces. This paper aims to demonstrate the value of using embeddings, \ie dense real-valued vectors representing each visited location, together with affordable methods to generate recommendations for the next Point of Interest (POI) to visit based on the last POI visited. In this study, we use the Word2Vec language model to generate these embeddings. Specifically, Word2Vec considers POIs as words and sequences of POIs as sentences. By associating a dense vector to each word, this model facilitates the capture of contextual information and enables the identification of words that share similar contexts, provided that their numerical vectors are close to each other. We will demonstrate that recommending the next POI based on Word2Vec embeddings is more beneficial compared with non-embedding classical methods. Empirical experiments conducted on a real dataset show that embedding-based methods outperform conventional methods in terms of prediction quality for recommending the next POI to visit.