Factors Influencing AI-Generated Code Quality: A Synthesis of Empirical Evidence

arXiv·medium signal

Geruslu, Aliyeva et al. synthesize empirical evidence across multiple studies on what factors actually determine the quality, reliability, and security of AI-generated code. While LLM coding tools promise productivity gains, the synthesis identifies specific conditions where AI code quality degrades — including prompt ambiguity, language-specific gaps, and security vulnerability patterns. Published in cs.SE/cs.AI, this is a practitioner-oriented evidence review rather than another benchmark paper.