Jia-Lei Chen published “Artificial Intelligence-Driven Software Test Automation: A Comprehensive Survey” in Applied and Computational Engineering on 2026-09-15, examining how AI can support software testing, a resource-heavy part of the software development lifecycle. The survey says traditional automated testing depends on predefined scripts and deterministic logic, which limits its ability to handle the dynamic complexity of modern software systems.

The paper reviews two key tasks: AI-driven test case generation and defect detection. It traces the shift from traditional automation to intelligent testing and analyzes methods including prompt engineering, retrieval-augmented generation (fetching outside information before answering), model fine-tuning and multi-agent systems (multiple AI agents coordinating tasks).

Chen contrasts traditional deep learning and large language model approaches for defect detection. The survey identifies hallucination, evaluation criteria, interpretability and generalisation as key challenges, and argues that large language models should move test automation from “execution automation” toward “decision support,” not replace human testers.

Jia-Lei Chen published “Artificial Intelligence-Driven Software Test Automation: A Comprehensive Survey” in Applied and Computational Engineering on 2026-09-15, examining how AI can support software testing, a resource-heavy part of the software development lifecycle. The survey says traditional automated testing depends on predefined scripts and deterministic logic, which limits its ability to handle the dynamic complexity of modern software systems.

The paper reviews two key tasks: AI-driven test case generation and defect detection. It traces the shift from traditional automation to intelligent testing and analyzes methods including prompt engineering, retrieval-augmented generation (fetching outside information before answering), model fine-tuning and multi-agent systems (multiple AI agents coordinating tasks).

Chen contrasts traditional deep learning and large language model approaches for defect detection. The survey identifies hallucination, evaluation criteria, interpretability and generalisation as key challenges, and argues that large language models should move test automation from “execution automation” toward “decision support,” not replace human testers.