Impact of Artificial Intelligence on Software Testing: A Comprehensive Research Manuscript

Abrar Hussain; Hamza Ali Kazmi; Muhammad Usman; Humaiz Shaikh; Faraz Ahmed Shaikh

Authors

Abrar Hussain Department of Software Engineering, ILMA University, Main Ibrahim Hyderi Road, Korangi Creek, Karachi, Pakistan
Hamza Ali Kazmi Department of Software Engineering, ILMA University, Main Ibrahim Hyderi Road, Korangi Creek, Karachi, Pakistan
Muhammad Usman Department of Software Engineering, ILMA University, Main Ibrahim Hyderi Road, Korangi Creek, Karachi, Pakistan
Humaiz Shaikh Department of Software Engineering, ILMA University, Main Ibrahim Hyderi Road, Korangi Creek, Karachi, Pakistan
Faraz Ahmed Shaikh Office of Quality Enhancement Cell, Dawood University of Engineering and Technology, Karachi, Sindh, Pakistan

Keywords:

Agentic Ai Testing, Automated Test Generation, Ci/Cd Quality Automation, Deep Learning, Defect Prediction, Fault Localization, Fuzz Testing, Large Language Models, Machine Learning, Metamorphic Testing, Quality Assurance, Reinforcement Learning, Self-Healing Automation, Software Testing

Abstract

The emergence and spread of various AI technologies have essentially transformed software testing from a manual, reactive process to an intelligent and predictive, and more autonomous one. This scholarly research paper reviews various effects that AI has on software testing – automated test cases generation, ML-based defects prediction, DL for GUI and fuzz testing, LLM-based QA, RL-based test exploration, metamorphic testing, fault localization, and self-healing CI/CD. The paper draws upon 93 peer-reviewed journal articles, IEEE and ACM conferences papers, and industry surveys and provides empirical evidence of benefits achieved through AI applications: 88% of enterprises use AI [1], 48% decrease in false positives [15], 40–60% increase in defect resolution speed [18], and the growth of testing market up to USD 284.73 billion by 2032 at 19.4% CAGR [57]. We provide original architectural diagrams, empirical performance bar charts, a radar comparison of traditional testing vs. AI-based, mathematical definitions of five important metrics of AI testing, a systematic literature review table of 20 studies, a comparative review table of 15 tools, and a challenge mitigation matrix of 10 items. In addition to issues such as transparency, data dependence, accountability, and oracular uncertainty, future research trends in agentic AI, neuro-symbolic inference, and federated learning are critically examined. The current manuscript is a compendium and evidence-based guide for AI in software quality engineering.