Generative AI in Software Engineering: Applications, Challenges, and Future Directions — A Survey

The rapid advancement of Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs) has significantly influenced modern Software Engineering practices. Tools based on models such as GPT-۴, Claude, Gemini, Code Llama, and GitHub Copilot are increasingly integrated into development environments to support activities across the Software Development Life Cycle (SDLC), including requirements engineering, design, code generation, testing, debugging, documentation, and maintenance. This paper presents a structured survey of recent research (primarily ۲۰۲۳–۲۰۲۵) on the applications of Generative AI in Software Engineering. Drawing on major systematic reviews and empirical studies, we synthesize key application areas, reported benefits, persistent challenges, and promising research directions. The evidence indicates that LLM-based tools improve productivity by automating repetitive tasks and reducing cognitive load, yet issues such as hallucinations, security risks, intellectual property concerns, prompt sensitivity, and the continuing need for human validation remain critical barriers to reliable industrial adoption. The future of Software Engineering is expected to center on effective Human–AI collaboration, supported by trustworthy AI techniques, domain-specific models, Retrieval-Augmented Generation (RAG), Explainable AI (XAI), and standardized evaluation frameworks.

کلیدواژه ها:

: Generative Artificial Intelligence ، Large Language Models ، Software Engineering ، Code Generation ، Software Testing