A Survey of AI-Driven Automated Game Testing Frameworks and Approaches

Abstract

Modern video games present significant testing challenges due to their complexity, non-deterministic systems, and emergent gameplay behaviour. Traditional manual and scripted testing approaches are often costly, time-consuming, and limited in scalability. This survey examines recent developments in AI-driven automated game testing, focusing on reinforcement learning, imitation learning, model-based testing, and hybrid exploratory approaches. Existing frameworks and tools are analysed in terms of adaptability, behavioural realism, scalability, and integration within modern development pipelines. The paper also identifies key research challenges, including generalisation, computational cost, explainability, and the lack of standardised evaluation methods. Finally, future directions for scalable and production-ready automated testing systems are discussed.