Using LLM-driven tools to review pull requests for bugs, style, and security—augments, does not replace, human review of non-trivial changes.
AI code review spans inline IDE feedback (GitHub Copilot Code Review) and PR-level bots (CodeRabbit, Qodo Merge). Gains depend on fit to repo conventions and review discipline.
Watch out
AI reviewers miss architectural and intent-level problems. Treat them as linters, not reviewers.