FOR THE BUILDERS, THE MAKERS & THE SMALL TEAMS WITH BIG IDEAS ✳
AI Coding · News Analysis

GitHub Copilot Adds Review Addressed Rates: What Small Teams Can Learn

GitHub’s November 5, 2026 update adds code review details and addressed rates to repository-level Copilot usage reports. Small teams can examine which comments prompt action and where feedback gets ignored. The most consequential detail is the definition: a comment can count as addressed when a developer promises a fix in a later pull request. Use the new figures to investigate review engagement, while checking correctness separately.

ai-coding: Official update → Team impact → Adoption decision. Editorial workflow illustration.
Source-based News Analysis. GitHub’s announcement is dated November 5, 2026; official sources accessed November 7, 2026. No authenticated reports were retrieved, API behavior tested, or review outcomes independently measured. Examples and evaluation steps are editorial proposals. Documentation may change.

1. Repository reports now follow review feedback

According to GitHub’s November 5 announcement, each enterprise and organization repos-1-day report row now includes a copilot_code_review object. It contains evaluated and addressed comment counts, breakdowns of those counts, and a count of merged pull requests without a human change request after Copilot’s first review.

This extends the repository reporting introduced in July. Earlier fields described pull request activity and applied suggestions. The new fields capture additional ways developers respond, including manual edits and requests for Copilot to make a fix.

For a small team already examining organization reports, this provides a more specific question to investigate: which review feedback receives attention? GitHub presents the update as a way to understand whether developers act on comments. The announcement supplies no independent evidence of improved code quality or reduced review effort.

2. Read the denominator before the percentage

GitHub defines total_comments_evaluated as the number of Copilot review comments checked for developer action. total_comments_addressed counts evaluated comments acted on through applying a suggestion, asking Copilot to fix it, editing code, or replying that a fix will follow in another pull request. The addressed rate is addressed comments divided by evaluated comments.

That final category matters. A recorded response can establish engagement while leaving the underlying issue unresolved. Similarly, an applied edit does not by itself establish that the resulting behavior is correct. Our AI code review workflow explains how to keep review findings connected to acceptance checks.

Our reporting recommendation is to display both counts beside the percentage. For illustration, three addressed comments out of four produce 75%, but that small sample offers little basis for a broad judgment. When the evaluated count is zero, show the rate as unavailable; division by zero cannot establish a zero-percent response rate.

3. Merge timing and bot activity affect comparisons

GitHub says comments are evaluated when their pull request merges. They appear in counts for the merge day, even if Copilot posted them earlier. A quiet review day can therefore receive counts from older work, while feedback on an open pull request has not yet entered these merge-based totals.

There is also a population difference. GitHub says it usually does not evaluate comments on pull requests opened by bots, including Copilot cloud agent. However, total_merged_no_human_change_request_after_review still includes those pull requests. Avoid assuming the comment counts and this pull request count describe identical populations.

The API documentation says repository reports include only repositories with activity on the requested day. Within returned rows, the announcement says repositories without Copilot code review activity have zero counts and empty arrays. A missing repository row and a present row with zero evaluated comments deserve separate handling in a report.

4. Use breakdowns to choose what to inspect

The update provides evaluated and addressed counts by severity, review tier, comment type, and customization. GitHub gives examples including high, medium, and low severity; lite, balanced, and max review tiers; and custom instructions, skills, MCP servers, or no customization.

These dimensions can help a team select a useful review sample. For example, inspect high-severity comments that received no recorded action, or examine whether comments associated with repository guidelines led to concrete changes. Those are proposed investigations, not observed results from this article.

A higher rate in one customization category cannot establish that the customization caused better feedback. Repositories may differ in task complexity, contributors, and review practices. Our coverage of customization usage metrics provides related context for interpreting customization activity. Keep the question narrow: does inspecting these comments reveal a specific instruction worth revising?

5. Start with one repository and inspect the outcomes

We recommend a bounded evaluation using one repository whose merged pull requests the team can inspect. Choose a reporting period, record evaluated and addressed totals, and retain the merge dates behind the sample. Calculate the period’s rate from summed counts; averaging daily percentages would give a quiet day the same weight as a busy one.

Then manually examine a manageable sample of feedback. Record whether the response involved a completed code change, a request for further work, or a deferred fix. Check whether the available diff and validation support the claimed resolution. These manual categories are an editorial proposal, not additional API fields.

Keep follow-up commitments visible in the team’s existing workflow. A comment already counted as addressed may still require implementation. Use the checks in our regression testing guide when a proposed fix changes behavior. We have not performed this evaluation or measured its results.

6. Make the metric support a review decision

The count of merged pull requests without a later human change request needs equally careful wording. Its definition records the absence of that particular event after Copilot’s first review. It does not establish that a person approved every change, that review was thorough, or that no defects remained.

Use the new reporting when your team wants to understand recurring feedback and can inspect the underlying work. Start with a concrete decision, such as whether a repository instruction produces useful comments or needs revision. Keep the counts, merge timing, and sample limitations attached to that decision.

If the report merely produces a percentage nobody investigates, defer further dashboard work. The strongest practical use is to direct attention toward comments and follow-up tasks that deserve a closer look. A rising addressed rate alone should not become a release acceptance criterion.

Sources & further reading

Primary references used for product and technical information. Features and terms can change; check the provider before committing.

No affiliate links are included in this article. Read our disclosure.