Generative AI for Software Engineering: A Review of Code Generation, Testing, Human–AI Collaboration, and Quality Assurance

Generative AI is increasingly reshaping software engineering (SE) across the development lifecycle. This review synthesises 113 peer-reviewed journal studies and organises the literature around five research questions and four application areas: code generation, testing and debugging, human–AI collaboration, and quality assurance. The reviewed studies show substantial progress in natural-language-to-code generation, context-aware programming, automated testing, fault localisation, program repair, and AI-assisted development. At the same time, persistent challenges remain in semantic correctness, non-deterministic outputs, software security, maintainability, verification, and evaluation consistency. Drawing these findings together, the paper presents a lifecycle mapping, a closed-loop AI-driven software engineering framework, and a four-dimension taxonomy to organise existing research and clarify connections across previously fragmented areas. Overall, generative AI is moving software engineering from a largely deterministic and human-centred workflow toward a more probabilistic and AI-enhanced engineering process. Further progress will depend on stronger verification, security, reliability, maintainability, and appropriately calibrated human oversight.