Current game worlds often fall short in providing consistency between the visual representation of the world and the way it feels, behaves, and reacts. This problem partly originates from the goaloriented and cost-effective nature of the game development process, which mostly favors ad hoc solutions for one particular game, rather than investing in concepts like reusability and emergent gameplay. In broader terms, we observe that game worlds miss semantics, and we argue that its deployment has the potential to bring about the consistency missing in their content. Therefore, we present a novel approach aimed at enriching virtual entities in game worlds with information about their roles, how they relate to others, and how they can affect and interact with players, NPCs, and with each other. We discuss several requirements to achieve these goals, and introduce a semantic model to represent game worlds. In order to support and validate this model, we have developed Entika, a framework to facilitate the deployment of semantics during game development, as well as its maintenance during run-time. Furthermore, we briefly discuss several applications that demonstrate the power of this semantic model for game worlds. After careful evaluation of our semantic game world model and framework, we conclude that a semantically rich world representation can substantially assist designers in creating much more consistent game worlds.