The Code Review Checklist for Test PRs (Most Teams Skip This)
TL;DR
A test PR deserves the same scrutiny as application code, because a bad test creates false confidence rather than just missing coverage. Check whether the test maps to a real journey, whether the assertion could pass while the feature is broken, and whether selectors and waits are resilient. Reject on sight: swallowed exceptions, assertions on implementation details, and copy-pasted tests with the assertion never replaced.
Application code gets a real review — logic checked, edge cases questioned, naming argued over. Test code gets an approve-on-sight because "it's just tests." That double standard is backwards: a bad test doesn't just fail to catch a bug, it actively creates false confidence that the bug was caught. A test PR that adds a check which can never fail is worse than no test at all, because it looks like coverage without being coverage.
Why Test PRs Deserve the Same Scrutiny
The cost of a bad test isn't paid at merge time — it's paid months later, when the feature it was supposedly guarding breaks in production and everyone assumed the green suite would have caught it. Reviewing test code with the same rigor as application code is cheaper than discovering the gap after a customer does.
The Checklist
- Does this test map to a real user journey, or a DOM node? "Verify the submit button is present" and "verify submitting the form creates the order" look similar in a diff and prove entirely different things.
- Could this assertion pass while the feature is broken? If the answer is yes for any plausible bug, the assertion is checking the wrong thing. This is the single most important question on the list — see why your app breaks when 100% of your tests pass for why it's so easy to get wrong without noticing.
- Are selectors semantic, not brittle?
getByRoleandgetByLabelsurvive UI refactors. A CSS path three divs deep does not. - Any hardcoded waits? A
waitForTimeoutin a new PR is a flaky test in waiting — see our reliability playbook for why condition-based waits are non-negotiable. - Is the test data isolated? New tests that reuse a shared seeded account are the most common source of parallel-run flakiness introduced after the fact.
- Does a failure point to one clear cause? A test that asserts five unrelated things in sequence fails for reasons that take longer to diagnose than the bug itself would have taken to fix.
Red Flags to Reject on Sight
- A broad
try/catcharound the test body that swallows failures instead of letting them fail loud. - Assertions on implementation details — an internal CSS class name, a Redux store shape — instead of user-visible outcomes.
- A new test copy-pasted from an existing one with the assertion deleted and never replaced.
Making Reviewers Actually Do This
The checklist only works if it's visible at review time, not buried in a wiki page nobody opens. Put it in the PR template for any diff touching the test directory. Pair it with the same journey map used elsewhere in your coverage process — see how to map critical user journeys — so reviewers have something concrete to check a new test against instead of relying on gut feel. Teams that skip this step tend to rediscover the per-test trap one bad PR at a time — see the per-test trap for what that pattern looks like at scale.
Want a second set of eyes on how your team currently reviews test PRs? Book a demo and we'll walk through it together.
Tags
See QA Guardian in action
Everything we write about is what we build and run every day. Book a demo and we'll show you on your own codebase.