Practise the language developers use when generating, reviewing, and discussing tests produced by AI tools — including how to talk about what AI gets right and what it misses.
0 / 5 completed
1 / 5
You prompt Copilot: '___ unit tests for the calculateTax function.' What are you asking the AI to do?
'Generate unit tests for' is a standard prompt pattern for AI-assisted testing. You provide the function (by naming it or pasting it) and the AI produces test cases covering happy paths, edge cases, and sometimes error conditions.
2 / 5
In a pull request description a developer writes: '___ generated these tests as a starting point — I've reviewed and added two extra edge cases.'
Stating 'Copilot generated these tests' in a PR description is transparent and professional — it tells reviewers the tests came from AI and that the developer has taken responsibility for reviewing them.
3 / 5
After reviewing AI-generated tests your colleague says: 'The AI ___ the edge case where the input array is empty.'
'The AI missed this edge case' is a common observation when reviewing AI-generated tests. AI tools often cover common paths well but may overlook boundary conditions, null inputs, or domain-specific scenarios the developer must add manually.
4 / 5
Your team notices that ___ coverage is 90% thanks to AI-generated tests, but a critical bug slipped through. What does this reveal?
High test coverage from AI-generated tests can create false confidence. Coverage measures which lines were executed, not whether the assertions are meaningful — AI tests can pass while missing the exact condition that causes a production bug.
5 / 5
A reviewer comments: 'These look like ___ tests — they pass the happy path but none of them check what happens when the API returns a 500.'
Shallow AI-generated tests cover the easy cases but miss error handling, unexpected inputs, and failure modes. Part of reviewing AI-generated tests is identifying and filling these gaps before the code reaches production.