Skip to content
Back to AI Coding

AI Code Review

Use AI to review diffs for bugs, smells, and missing tests before human review.

Testing & ReviewIntermediate

Why

GitHub's 2024 survey found 92% of US respondents use AI coding tools for test case generation, and 98% of organizations have experimented with AI for tests. AI review catches mechanical issues (missing tests, banned patterns, edge cases) before humans, so human review focuses on architecture and intent.

It scales with PR volume.

When

Use on every PR before human review. Skip for trivial doc-only PRs.

How

1. Feed the diff + relevant context files. 2. Ask: "review for bugs, edge cases, missing tests, banned patterns; list concrete issues". 3. Reject vague feedback ("consider improving") — demand line + issue + fix. 4. Fix real issues; dismiss false positives with a reason. 5. Track AI vs human agreement to tune.

Example

Diff: +10 lines in books.ts (new route, no validation, no test) AI review: - books.ts:42 no input validation — add zod schema (bug). - no test for POST /books — add one (missing test). - uses 'any' for body — type with BookInput (banned pattern). Fix all three; dismiss "consider a service layer" as out of scope.

Common Problems

  • ⚠Error: Cannot find module 'jest' / 'vitest' → run `npm install --save-dev vitest` (or jest); verify the test script in package.json.
  • ⚠Error: "Test timed out after 5000ms" → increase `testTimeout` in config or await the async operation; common with network calls.
  • ⚠AssertionError: expected undefined to equal 'x' → the mock returned undefined; verify the mock setup returns the expected shape.

Best Practices

  • ✓Require concrete, line-referenced feedback with a suggested fix.
  • ✓Log dismissed false positives to prune noisy rules over time.
  • ✓Measure the outcome with a concrete metric (time saved, defect rate, test coverage) so the practice can be justified and improved.

FAQ

What is AI Code Review?

Use AI to review diffs for bugs, smells, and missing tests before human review. It helps developers quickly grasp the core value and use cases of AI Code Review.

How do I get started with AI Code Review?

To get started with AI Code Review, focus on these points first: Feed the diff + relevant context files.

What are common mistakes with AI Code Review?

Common mistakes when using AI Code Review include: Error: Cannot find module 'jest' / 'vitest' → run `npm install --save-dev vitest` (or jest); verify the test script in package.json.; Error: "Test timed out after 5000ms" → increase `testTimeout` in config or await the async operation; common with network calls.; AssertionError: expected undefined to equal 'x' → the mock returned undefined; verify the mock setup returns the expected shape..

What are the best practices for AI Code Review?

Best practices for AI Code Review include: Require concrete, line-referenced feedback with a suggested fix.; Log dismissed false positives to prune noisy rules over time.; Measure the outcome with a concrete metric (time saved, defect rate, test coverage) so the practice can be justified and improved..

How do I learn more about AI Code Review?

Start with the official docs, then search real-world examples on GitHub. Validate patterns in a small scratch repo before applying them to production code.