In today’s competitive media landscape, effectively pitching press releases to the right journalists is key to maximizing engagement and product reach. To address this challenge, this study proposes a hybrid framework combining topic classification, recommendation, and text similarity methods. Using text processing techniques, semantic features are extracted from a large corpus of news articles by leveraging GloVe embeddings. A supervised SVM categorizes pitches into five topics, narrowing the pool of potential journalists. A nearest-neighbor search in the vector space identifies the most similar articles on the same topic to generate a list of recommended journalists. Various accuracy, ranking, coverage, and diversity metrics evaluate the system’s performance, which achieves the right balance between relevant and novel recommendations.