This survey explores the transformative impact of foundation models and learned game-world models across the entire game development lifecycle. The authors categorize current AI applications into six distinct functional roles: playing and acting, modeling players and games, designing games, building and maintaining games, generating and adapting content at runtime, and testing and evaluating games. The research highlights critical cross-role synergies, such as using player trajectories to train world models, leveraging learned environments to provide experience for agents, and utilizing design specifications to drive executable implementations. Despite these connections, the authors note that many AI capabilities remain constrained by setting-specific variables, including control schemes, engine interfaces, state representations, and player contexts. While evaluation metrics are well-standardized for bounded game playing and specific learned environments, the paper identifies significant gaps in establishing rigorous evidence for more complex tasks. Specifically, more robust validation is required for maintaining persistent state in learned worlds, automated software revision, validated player modeling, sustained runtime adaptation, and representative automated testing. The study concludes that the primary technical challenge lies in developing methods to transfer outputs and capabilities across different roles while ensuring effectiveness within specific game-driven contexts.