This paper explores how interaction with large language models (LLMs) can lead to emergent behaviors in game narratives, allowing players to shape the story. The study uses a text-adventure game called Dejaboom! where players attempt to solve a mystery by interacting with non-player characters (NPCs) generated by GPT-4. The game is implemented using TextWorld, with GPT-4 providing dynamic dialogue with NPCs. Players are given the opportunity to explore and interact with the game in a non-scripted manner, leading to the creation of new narrative paths not originally planned by the game designers. The study involved 28 gamers who played the game and provided feedback. The researchers analyzed the game logs to create narrative graphs, identifying emergent nodes—new narrative paths created by players. These emergent nodes often involved creative strategies for interacting with NPCs, discovering new objects or locations, and even new ways to defuse the bomb. Players who created the most emergent nodes were those who enjoyed games that encouraged discovery, exploration, and experimentation. The study found that emergent nodes can lead to new, creative gameplay experiences, even if they do not always result in a successful outcome. Players appreciated the flexibility and creativity of the game, despite some frustrations with latency and NPC consistency. The research suggests that LLMs can be used as a tool to enable player-driven narrative development, allowing for more dynamic and engaging game experiences. The findings indicate that players with a creative motivation profile are particularly suited to this type of player-driven emergence, potentially leading to a more collaborative approach to game design involving players, designers, and LLMs. The study highlights the potential of LLMs in creating more open and dynamic game narratives, offering new opportunities for player creativity and engagement.This paper explores how interaction with large language models (LLMs) can lead to emergent behaviors in game narratives, allowing players to shape the story. The study uses a text-adventure game called Dejaboom! where players attempt to solve a mystery by interacting with non-player characters (NPCs) generated by GPT-4. The game is implemented using TextWorld, with GPT-4 providing dynamic dialogue with NPCs. Players are given the opportunity to explore and interact with the game in a non-scripted manner, leading to the creation of new narrative paths not originally planned by the game designers. The study involved 28 gamers who played the game and provided feedback. The researchers analyzed the game logs to create narrative graphs, identifying emergent nodes—new narrative paths created by players. These emergent nodes often involved creative strategies for interacting with NPCs, discovering new objects or locations, and even new ways to defuse the bomb. Players who created the most emergent nodes were those who enjoyed games that encouraged discovery, exploration, and experimentation. The study found that emergent nodes can lead to new, creative gameplay experiences, even if they do not always result in a successful outcome. Players appreciated the flexibility and creativity of the game, despite some frustrations with latency and NPC consistency. The research suggests that LLMs can be used as a tool to enable player-driven narrative development, allowing for more dynamic and engaging game experiences. The findings indicate that players with a creative motivation profile are particularly suited to this type of player-driven emergence, potentially leading to a more collaborative approach to game design involving players, designers, and LLMs. The study highlights the potential of LLMs in creating more open and dynamic game narratives, offering new opportunities for player creativity and engagement.