GitHub Copilot code review: request, automate and customize reviews

GitHub Copilot code review is Copilot acting as a reviewer on your pull requests: you request it like a teammate (or set it to review automatically), and it leaves comments labeled High, Medium or Low severity, often with suggested changes you can commit in a couple of clicks. It also reviews selected code or uncommitted changes in your IDE.

GitHub's docs describe it as a feature that "reviews your pull requests, identifies issues, and suggests fixes you can apply in a couple of clicks." This guide is based on GitHub's Copilot code review documentation read on 24 October 2026.

Where can you use Copilot code review?

GitHub lists support in GitHub.com, GitHub CLI, GitHub Mobile, VS Code, Visual Studio, Xcode, JetBrains IDEs and Azure DevOps (public preview).

Plan support, from GitHub's plans page:

If you get Copilot from an organization, an admin must enable the Copilot code review policy for reviews on GitHub.com and GitHub Mobile. Business and Enterprise organizations can also let members without a Copilot license use code review on GitHub.com; that needs two policies, AI credits paid usage and Allow members without a Copilot license to use Copilot code review in GitHub.com, and the usage is billed to the organization.

How do you request a Copilot code review?

On GitHub.com, from GitHub's using code review guide:

1. Create or open a pull request.
2. Under Reviewers in the right sidebar, click Request next to Copilot.
3. Wait. GitHub says this "usually takes less than 30 seconds."
4. Read the comments. Each has a severity label of High, Medium or Low.
5. Commit a single suggestion, or batch several into one commit.

From the GitHub CLI:

# request Copilot when opening a pull request
gh pr create --reviewer @copilot

# add Copilot to an existing pull request
gh pr edit PR-NUMBER --add-reviewer @copilot

You can also request copilot-pull-request-reviewer[bot] as a reviewer through the REST API.

Copilot's comments behave like a human's: you can react, reply, resolve and hide them. One catch: replies you add are visible to people but "won't be visible to Copilot, and Copilot won't reply." To have the fixes made for you, click Fix with Copilot on a comment, which hands the work to Copilot cloud agent (passing suggestions to cloud agent is in public preview) and can open a new pull request against your branch or commit to the same one.

How do you review code in your IDE?

In VS Code you have two options:

• Review a selection. Select code, right-click and choose Generate Code > Review. Comments appear in the Comments panel and inline.
• Review uncommitted changes. In the Source Control view, hover over CHANGES and click the Copilot Code Review - Uncommitted Changes button. Comments appear inline and in the Problems tab.

Reviewing before you push catches issues while they are still cheap to fix.

How do you set up automatic reviews?

By default Copilot only reviews when asked. GitHub's configuration guide covers three levels:

For your own pull requests. Profile picture > Copilot settings > Code review, then enable Automatic Copilot code review. This needs Pro, Pro+, Max, or a Business or Enterprise license, and isn't available for managed user accounts.

For a repository. Go to Settings, then Rulesets under "Code, planning, and automation", click New ruleset > New branch ruleset, set it to Active, add target branches and select Automatically request Copilot code review.

For an organization. Organization owners can do the same with organization rulesets across some or all repositories.

Two options change when reviews run:

• Review new pushes re-reviews on every commit. Without it, Copilot reviews a pull request once.
• Review draft pull requests reviews drafts before they are marked ready.

Your user settings and rulesets are evaluated separately. If both apply, Copilot still posts only one review, and your personal settings can't switch off push or draft reviews a ruleset turned on. To re-review manually, click the re-request button next to Copilot in the Reviewers menu. GitHub notes that a re-review may repeat comments you already resolved or downvoted.

What are Lite and Balanced review effort?

GitHub offers two effort levels:

• Lite: fast, targeted feedback on common issues "such as bugs, security vulnerabilities, and style inconsistencies."
• Balanced: routes the pull request "to a higher-reasoning model for longer analysis of complex logic, security-sensitive code, and cross-service changes." This is the built-in default.

A Max level appears in personal settings marked Coming soon and is not available yet.

You can pick the effort under Reviewers when requesting a review, or set defaults in your personal settings, repository settings or organization settings. GitHub recommends Balanced for security-sensitive code and multi-service changes, and Lite for routine changes where speed matters more. GitHub changed the default from Lite to Balanced on 28 September 2026, per its August changelog.

You cannot pick the model. GitHub says "model switching is not supported," because the review uses a tuned mix of models and prompts.

How do you customize what Copilot looks for?

Copilot code review reads the same instruction files as other Copilot agents:

• .github/copilot-instructions.md for repository-wide review rules.
• .github/instructions/**/*.instructions.md for rules that apply only to matching paths.
• AGENTS.md in the repository root for context shared across AI tools.
• CLAUDE.md, GEMINI.md and REVIEW.md, if they exist.

GitHub's example instructions file:

When performing a code review, respond in Spanish.

When performing a code review, apply the checks in the `/security/security-checklist.md` file.

When performing a code review, focus on readability and avoid nested ternary operators.

Copilot reads instructions and skills from the head branch, so you can test changes to them in the same pull request. See the custom instructions guide for file formats.

Beyond instructions:

• Agent skills in .github/skills are used when relevant. To make sure a skill is picked up for reviews, GitHub suggests a review-focused directory name such as code-review. See the Copilot agent skills guide.
• MCP servers configured in repository settings can pull in context from issue trackers or incident tools. The GitHub MCP server and Playwright MCP server are on by default.
• Environment. Reviews run in an ephemeral environment. Use .github/workflows/copilot-code-review.yml to configure it separately, or it falls back to the cloud agent's copilot-setup-steps.yml.

Some files are never reviewed: dependency files such as package.json and Gemfile.lock, log files and SVG files.

Can Copilot approve pull requests?

Not by default. Copilot leaves a "Comment" review, which doesn't count toward required approvals. Every review includes an approval assessment in its overview comment, though.

Copilot approvals are in public preview. When enabled, Copilot "can submit an approving review that satisfies your repository's required-approval rule the same way a teammate's approval would." New commits dismiss the approval. In repository settings, you can limit which pull requests count toward merge requirements with up to 15 file path globs.

GitHub's advice still applies: "Copilot is not guaranteed to spot all problems or issues in a pull request." Keep a human reviewer for anything that matters.

How much does Copilot code review cost?

Each review has two costs:

1. AI credits for the model work. GitHub estimates "$0.05 USD to $1 USD worth of AI credits with 'Lite' effort, and $0.25 USD to $5 USD worth of AI credits with 'Balanced' effort." Bigger pull requests and longer custom instructions cost more.
2. GitHub Actions minutes for the agentic part (full project context gathering and handing fixes to the cloud agent). Per GitHub's April changelog, this started on 1 June 2026 for private repositories; Actions minutes on public repositories remain free. Self-hosted runners don't consume Actions minutes.

Automatic reviews are billed to the pull request author by default, and manual requests to whoever requested. For how credits work across plans, see the GitHub Copilot pricing guide.

Keep your review checklists reusable

The review rules that catch real bugs in your codebase are worth writing down once. Kipt keeps prompts and Agent Skills, such as a code-review skill, as Markdown in a private GitHub repo you own, so you can reuse them across projects. For a ready-made starting point, see the code review skill and these code review prompts.