Many developers use GenAI like GitHub Copilot to generate their unit tests. For most developers, having just decent code coverage is enough, and the quality of the unit tests relative to the test theory is not examined. In this talk, we will, however, do just that.

Based on some working code, we will generate unit tests and check the generated tests for compliance with the test theory. Do not expect a lot of code. We will spend the better part of the talk in code.

In the session, we will use C#, xUnit and GitHub Copilot/JetBrains Junie, but the information is usable across multiple programming languages and agentic coding tools.

At the end of the session, you will leave with enough knowledge to identify the gap between the generated test code and the test theory, and have some insights into how to steer the agentic coding tools to generate better test code.