Recommender systems (RSs) are tools that help users to make decisions, when they face overloaded domains, suggesting the best items according to their tastes and preferences. Content-based RSs (CBRSs) are one of the most popular types of RSs, which work with items’ descriptions and users’ profiles to make recommendations. Besides, group recommendation systems (GRSs) are another type of RS which are focused on user groups rather than individual users, providing recommendations based on the group preferences. The present work proposes a new method of content-based group recommendations (CB-GRSs), which represents a line of work scarcely addressed by the scientific community, with three main novelties: weighting of item features, integrating group profiles in the set of user profiles and selecting dynamically the best aggregation function depending on the size of the group. An experimental study is carried out to show the advantages of the proposal with respect to previous works.