A crucial component of storytelling in the now 200 billion-dollar gaming industry, Non-Playable Characters (NPCs) in video games often exhibit scripted behaviors and limited interactivity, which greatly prevents proper player immersion. As LLMs grow in popularity and usage across various fields, this study aims to explore the integration of Large Language Models (LLMs) in Unreal Engine 5 using Convai’s API to develop dynamic, lifelike NPCs with real-time, context-aware interactions. With the implementation running on computers equipped with an RTX 3050 GPU and 32GB RAM, results indicated a significant increase in engagement by 10–30 percent, along with an improvement in NPC realism. The use of LLMs would also allow game developers to allocate more time to optimization and fine-tuning of game mechanics, further enhancing the player’s experience. This study demonstrates the viability and effectiveness of LLM-based NPCs, as well as analyzing their benefits and drawbacks.