Monkey++ AI boosts Android UI testing 10x with smarter flows

A team of researchers from University College Dublin and the National Institute of Informatics has developed Monkey++, an AI-powered enhancement to Android's decades-old Exerciser Monkey testing tool. The core innovation is a novel control flow structure that models the interactive elements of mobile apps, giving the tester semantic awareness of the application's UI architecture. Unlike traditional monkey testing—which relies on random inputs or tedious hand-crafted scenarios—Monkey++ uses this structure to guide its exploration with purposeful depth-first search (DFS), eliminating redundant tests and focusing on unvisited paths. In benchmarks, Monkey++ achieved full coverage of interactive UI elements 10x faster than the baseline Exerciser Monkey, transforming a process that previously took hours into one that completes in minutes.

The paper, published on arXiv (arXiv:2608.00236), argues that traditional UI testing tools are blind to application logic and thus inefficient. By modeling the app’s state space and user interaction graph, Monkey++ not only accelerates testing but improves its reliability. The authors suggest this approach could lead to a new generation of model-driven testing tools across platforms, reducing manual QA effort and enabling continuous integration pipelines to catch UI/UX bugs earlier in development cycles.