Recommender systems are used to recommend books to users that are similar to their interests by analyzing behavioral patterns and preferences. However, traditional methods like collaborative filtering and content-based methods fail in some cases, such as the cold-start problem and data sparsity. To enhance accuracy and diversity, we proposed a hybrid book recommendation system by combining popularity-based filtering, collaborative filtering, and content-based filtering. This model leverages user ratings, general book popularity, and semantic similarities derived from book descriptions using Sentence BERT embeddings. For evaluating, we used the Goodreads dataset, which includes a wide range of English-language user reviews. Experiments compare multiple strategies, including user-based and item-based collaborative filtering, matrix factorization (SVD), and semantic content embedding. Results demonstrate that combining popularity and content features with collaborative methods improves recommendation relevance and robustness. The proposed hybrid model achieves higher accuracy and scalability, offering a well-rounded reading experience for a diverse global audience.