In recent years, the proliferation of natural language processing technologies has profoundly reshaped human-computer interactions. A notable example is OpenAI’s ChatGPT, a large language model (LLM) that has gathered attention worldwide across diverse fields. This study delves into the intriguing convergence of LLM and collaborative play within tabletop role-playing games (TTRPGs). To comprehend this interplay, I employ Actor-Network Theory (ANT) as the conceptual framework, enabling a nuanced analysis of this complex interaction. In this study the LLM, particulartly OpenAI’s ChatGPT-3.5, assumes multifaceted roles during gameplay: in one istance ChatGPT functions as the player who contribute to the evolving story, and simulate characters; in another instance ChatGPT takes the role as game master who orchestrate events, and challenges. The first seed of each instance was generated by the author of this study, afterwards the human role was a merely facilitator of the two instances, aka copying and pasting the outputs in one to another.Furthermore, drawing inspiration from Algirdas Julien Greimas’s semiotic theories, particularly actant schema and narrative program the basis of ANT, I could exame the intricate relationships between the two instances of LLM and the game system itself, to finally conceptualize LLMs as active participants rather than mere tools, shaping the narrative fabric of gameplay. The performance blur traditional boundaries, challenging established notions of agency and authorship. This study extends beyond mechanics and might contribute to the ongoing discourse on play, agency, and AI-driven creativity.