DragonCrawl: A Generative, Intent-Based Framework for Scalable Mobile End-to-End Testing

DragonCrawl is an AI-driven mobile regression-testing system that evolved from embedding-based UI similarity matching to generative, intent-based reasoning with GPT-4o’s multimodal capabilities. It validates specified user flows on every code change and can block commits that break critical functionality. The paper reports 91.6% and 92.2% pass rates on iOS and Android across 1,013 automated tests, respectively. It also claims that onboarding time fell from 96–120 hours to under four hours, while production deployment saved an estimated 27 developer-years of test-maintenance effort. The architecture combines visual end-state detection with tool calls for backend state transitions.