Procedural Content Generation (PCG) enables games to produce vast amounts of content through algorithmic systems, yet large-scale generation often suffers from a persistent novelty problem in which content becomes structurally repetitive despite high combinatorial variety. This paper examines the limitations of traditional PCG techniques and analyzes how Novelty Search and Neuroevolution can sustain long-term diversity by evolving content generators rather than optimizing toward a fixed objective. Using level generation as a case study, we describe how novelty-driven evolution promotes unique structural patterns and explore the gap between algorithmic diversity and player-perceived novelty. The results highlight both the promise and the challenges of using evolutionary, novelty-centered approaches to produce content that remains engaging, surprising, and meaningful over extended periods of play.