The integration of Large Language Models (LLMs) into game development presents an exciting frontier with the potential to revolutionize player experiences. LLMs offer unprecedented capabilities for generating dynamic content, creating believable and reactive characters, and enabling novel forms of player interaction. This project aims to explore the technical aspects of incorporating LLMs into games, focusing on their potential to enhance gameplay through dynamic content generation, AI-driven characters, and novel interaction paradigms. This project will investigate the practical application and technical challenges of LLMs in creating richer and more engaging game worlds. It will primarily focus on the technical implementation and evaluation of LLM integration, rather than game design or artistic aspects. Abstract: This project focuses the development of a Large Language Model (LLM)-driven Non-Player Character (NPC) within a dialogue-based game. The project takes the form of a single-player role-play scenario, in which the LLM assumes the role of a terrorist while the player acts as a counterterrorist tasked with interrogating the NPC to discover the location of a planted bomb. Unlike traditional scripted NPCs, the LLM-powered character demonstrates dynamic, adapting its responses to player input and maintaining an interactive, evolving dialogue. After applying various techniques and conducting a series of preliminary tests, a user study was carried out to evaluate whether the LLM-powered NPC genuinely enhanced the gaming experience. 91.7% of the participants agreed that the LLM-powered NPC enhanced the gaming experience, particularly by introducing variability, promoting problem-solving and creative thinking, and presenting challenges to the players. Based on the completed post-survey responses and the conversational data exchanged between players and the NPC, several adjustments were made to the game before conducting the comparative analysis across the three LLMs. The findings from this research demonstrate that LLM-driven NPC can move beyond novelty to deliver adaptive, engaging, and challenging gameplay, while also highlighting important directions for improving realism, memory, and consistency in future game design.