A study by the University of Maryland and Google DeepMind introduces StoryScope, an automated pipeline detecting AI-written texts with 93.2% accuracy based on narrative features. Analyzing over 60,000 stories from humans and five LLMs (Claude, DeepSeek, Gemini, GPT, Kimi), researchers found AI exhibits distinct narrative fingerprints: overly didactic, linear plots with few subplots, excessive physical descriptions, and no reader interaction. Each model has unique flaws. The code and text library are publicly available.