Machine Learning Approaches for Auto-Repairing Test Steps in UI Automation: A Review

Abstract

In modern software development, UI automation testing is one of the most important aspects of application quality assurance. Excessive UI changes make automated test scripts fragile, hence resulting in flaky tests and exorbitant maintenance costs that expose QA teams to unbearable stress. To overcome such issues, recent progress in Machine Learning (ML) and Artificial Intelligence (AI) has been leveraged to develop auto-repair mechanisms for automatically detecting and repairing broken test steps, making tests more reliable and reducing human effort. This survey presents an overview of current ML-based solutions to self-healing UI test automation, including element locator repair, flaky test stabilization, and wait automation. We address methods from heuristic to large language model-based repair and dynamic web and mobile test script modification. Leading frameworks describe how the integration of ML can make test maintenance easier and the tests themselves more resilient. Our presentation recognizes that while locator repair approaches have matured and offer immediate practical advantages, newer LLM-based approaches offer improved semantic understanding but at high costs of explainability and computational complexity. We talk about current constraints on the generalizability and explainability of test repair and provide guidelines for future research on advancing intelligent, adaptive QA automation solutions that reduce downtime and enhance software delivery quality