APPLICATION OF ARTIFICIAL INTELLIGENCE METHODS FOR AUTOMATED WEB APPLICATIONS TESTING

Abstract

Modern single-page web applications (SPAs) developed using React, Angular, and similar JavaScript frameworks are characterized by dynamic interfaces, asynchronous interactions, and continuously changing Document Object Model (DOM) structures. These features significantly complicate software quality assurance and reduce the effectiveness of traditional automated testing approaches based on static selectors and manually created test scripts. This paper investigates the application of artificial intelligence methods to improve the automation of functional and visual testing of modern web applications. A hybrid testing approach is proposed that integrates the capabilities of Large Language Models (LLMs) for intelligent test case generation with Computer Vision techniques for automated visual interface validation. The proposed framework automatically transforms natural-language functional requirements and user stories into executable Playwright test scripts while simultaneously performing visual regression analysis using OpenCV-based image comparison algorithms. To improve the robustness of automated testing, the framework incorporates mechanisms for selector self-healing, semantic analysis of detected visual differences, and adaptation to dynamic DOM modifications typical of SPA architectures. The technical implementation of the framework is based on Python 3.11 and integrates Playwright, OpenAI API, OpenCV, scikit-image, and asynchronous execution mechanisms, enabling seamless integration into modern CI/CD pipelines.

The effectiveness of the proposed approach was experimentally evaluated using three representative single-page applications developed with React, Angular, and Vue.js. Each application included multiple functional scenarios covering user authentication, data processing, shopping cart operations, dashboard management, and dynamic user interactions. Experimental results demonstrate that the hybrid AI-based framework successfully detects 93.3% of functional defects and 86.7% of visual regressions while significantly reducing the number of false-positive detections compared with conventional snapshot testing techniques. Furthermore, the proposed solution decreases automated test development time by more than four times and substantially reduces maintenance effort through intelligent adaptation to interface modifications and automatic selector recovery. The obtained results confirm that combining LLM-driven reasoning with computer vision techniques enables more reliable, scalable, and maintainable automated testing for highly dynamic web applications. The proposed methodology contributes to the development of next-generation AI-assisted software quality assurance systems and demonstrates strong potential for practical implementation in industrial software engineering projects involving continuous integration and continuous delivery environments.