Review AI-generated code before merging by assuming it is an untrusted contribution that must prove the requested behavior. Start with the requirement and acceptance criteria, then inspect the plan, diff, tests, command results, and unresolved risks. Do not begin with formatting details.
Confirm that the change solves the stated problem and does not expand scope. Review public interfaces, authorization, input validation, data writes, migrations, concurrency, error handling, logging, dependencies, and configuration. Trace important callers rather than reading only changed lines. Run tests and builds independently in a clean environment; add a test for the failure or behavior the change claims to address. Check that tests can fail for the right reason and that generated snapshots or fixtures were not accepted blindly. Inspect deleted code and permissions as carefully as additions.
Ask whether another maintainer can understand and reverse the change. Require a rollback or containment path for risky releases. Separate unrelated refactors and dependency upgrades so their effects remain reviewable. Treat confident explanations as navigation aids, not evidence.
Merge only when code, tests, and documented intent agree. For security, billing, identity, destructive data changes, or unfamiliar architecture, get qualified review. AI can summarize and find candidates for attention, while the named human reviewer remains responsible for the decision.
