Manual testing, where testers follow natural language instructions to validate system behavior, remains a critical part of software quality assurance. According to a new arXiv paper, such testing is still essential for uncovering issues that are difficult to capture with automation. However, the paper notes that manual test cases often contain problems—referred to as test smells—that can undermine their effectiveness.
The study, titled "An Empirical Study of Gemini 3 for Detecting Natural Language Test Smells in Manual Test Cases," examines how well the Gemini 3 model can identify these smells. The authors position this as an empirical investigation, meaning the work is based on observed performance rather than theoretical speculation. The abstract does not disclose specific results, so the paper's conclusions are not yet available from the provided text.
Because this is a single source, there are no conflicting findings to compare. The significance lies in exploring whether large language models like Gemini 3 can assist in improving manual test case quality, a task that has traditionally relied on human review. Further details on methodology and outcomes would require reading the full paper.