As software systems grow in complexity—especially in cloud-based microservice architectures, automated testing has become crucial for reliability and security. While academic REST API fuzzing tools (e.g., EvoMaster, Schemathesis) show strong fault detection, their adoption in industry is limited due to insufficient human-centered evaluation focusing on usability, learnability, and integration. This PhD project will address the gap through a mixed-methods, human-centered evaluation approach. Phase 1 will review current empirical methods in automated testing. Phase 2 will run controlled lab studies with students to evaluate tool usability and cognitive load. Phase 3 will bring the evaluation to real-world industrial settings. By triangulating technical performance with cognitive, behavioral, and perceptual data, this project aims to foster more usable and effective testing tools. It will contribute empirical methods and design recommendations to align academic advancements with real-world development needs.