An Empirical Study of Generative AI for CRUD Code: A PHP Case with Tutorial Based Novice Baselines

Abstract

The rise of generative artificial intelligence is changing the way software is developed, particularly in automating code generation. ChatGPT is one of the AI tools capable of producing executable code from natural language instructions. However, little is known about how its output compares to code written by beginners following online tutorials. This study explores ChatGPT's ability to generate a PHP based CRUD (Create, Read, Update, Delete) application and evaluates its quality against novice implementations from popular resources like Tutorial Republic. Using five criteria from ISO/IEC 25010:2011 readability, efficiency, logical structure, basic security, and maintainability. The study finds that ChatGPT consistently produces code that is more readable, less complex, and easier to maintain (18% vs. 8% comments, 2.5 vs. 4.2 cyclomatic complexity, 78 vs. 62 maintainability index). These results suggest that ChatGPT can generate functional, maintainable PHP CRUD code at or above the level of typical beginners. The study highlights the promise of generative AI as a practical tool to support programming education and early stage software development

Ähnliche Arbeiten