Genre, Bias, and Narrative Logic in AI Dungeon: Generative AI as a Game-Based Storytelling Engine

Objectives. This article investigates how generative artificial intelligence (AI) systems, as implemented in the digital game AI Dungeon, negotiate genre conventions and algorithmic bias within interactive, player-driven storytelling. By focusing on AI Dungeon as both a game and a narrative engine, the study aims to reveal how generative AI shapes narrative logic, aesthetic adaptation, and cultural representation in digital game environments.
Methodology. The research draws on a systematically coded dataset of AI Dungeon play sessions across multiple genres. Integrating theoretical perspectives from game studies, narratology, and algorithmic bias research, the study employs a mixed-methods approach. Quantitative coding captures genre, bias, narrative drift, and player intervention, while qualitative thematic analysis explores aesthetic adaptation and the negotiation of meaning between player and AI.
Findings. Results demonstrate that genre conventions in digital games serve as both scaffolds and filters for AI-generated storytelling. While genre prompts can guide the AI toward coherent and engaging narratives, they also interact with and sometimes amplify underlying algorithmic biases. Instances of stereotype reinforcement, narrative drift, and the need for player intervention are shaped by both genre and the ludic structure of the game.
Value. By empirically mapping the interplay between genre, bias, and interactive narrative logic in a digital game context, this study advances understanding of generative AI’s creative and ethical challenges in contemporary storytelling. The findings offer actionable insights for both game designers and AI narrative system developers.