This paper presents Star Quizller, an AI-powered trivia game designed to address the limitations of traditional trivia experiences that rely on static question banks and quickly lose replay and educational value. Leveraging OpenAI’s Large Language Models (LLMs), Star Quizller dynamically generates a limitless variety of trivia questions across multiple domains, ensuring each session is unique and tailored to player performance. The game integrates real-time procedural content generation, adaptive difficulty, and detailed performance analytics to enhance engagement, foster sustained curiosity, and promote enhanced learning. A mixed-method user study with 20 participants demonstrated significant improvements in knowledge retention, high replayability, and strong user satisfaction with the game’s educational and entertainment value. Our initial empirical results highlight the effectiveness of using AI-driven content and personalized feedback and allow further refinement in prompt engineering. Star Quizller exemplifies the potential of AI-enhanced educational games to deliver engaging, adaptive, and cognitively enriching experiences for diverse audiences.