Skip to main content

Improving Test Coverage with AI

AI can help you move from low coverage to meaningful coverage quickly. The goal isn't to hit a percentage; it's to cover the scenarios that matter.

Find uncovered paths​

Run your coverage tool and paste the report or the uncovered code:

Here is my coverage report. These lines are uncovered:

src/utils/permissions.ts: lines 34-41, 67-72

Here's the file: [paste]

Write tests that cover the uncovered branches.

Identify high-risk untested code​

Rather than chasing 100% coverage mechanically, ask the model to prioritise:

Here are the files with the lowest test coverage in our codebase:

[list files]

Which of these are highest risk if they contain bugs? Rank them and explain why.

Generate property-based tests​

For pure functions, property-based testing finds edge cases no one thought of:

Write property-based tests for this sorting function using fast-check.
Define properties that should hold for any valid input array.

[paste function]

Fill coverage gaps systematically​

Work file by file. For each uncovered area:

  1. Ask the model to explain what the code does
  2. Ask it to list all possible paths through the code
  3. Ask it to write a test for each path
Explain what each branch of this switch statement does,
then write one test per branch.

[paste code]

Coverage anti-patterns to avoid​

When using AI for coverage, watch out for:

  • Tests that only assert the call happened, not that the result is correct
  • Mocking everything: tests that mock so much they don't test real behaviour
  • Duplicate tests: AI sometimes generates near-identical tests for the same path
  • Tests with no assertions: always verify there's at least one expect() per test

If you need to write tests from scratch rather than fill coverage gaps, start with Writing Unit Tests with AI. For techniques on writing better prompts for test generation, see Advanced Prompting.