SNACK in three lines
- GitHub announced analysis-depth and efficiency updates for Copilot code review on June 25. Organizations in the Medium review effort level public preview can now see a Medium label in PR comments and set a default review level at the organization level.
- The back-end change may matter even more. Copilot code review now uses the grep, rg, glob, and view file-navigation tools from Copilot CLI and SDK to find the files it needs during review.
- GitHub says this adjustment reduced code review costs by about 20% while maintaining its review-quality bar. In other words, beyond simply saying “AI reviews code,” teams now need to manage review depth, cost, and defaults together.
Snackgirls editor note
AIKO: “This time, the point is less about a flashy new model and more about how teams control AI reviews at an organizational level. Making review depth visible and letting teams set defaults is an operations-side change.”
Red: “The good point is the 20% figure. Instead of a vague claim that it got smarter, there is a number showing review costs went down, so development teams can compare it right away.”
What changed
GitHub announced analysis-depth and efficiency updates for Copilot code review on June 25. Teams in the Medium review effort level public preview can now check in the Pull Request overview comment whether a review was generated with Medium analysis depth.
Organization administrators can also set a default organization review level for repositories that do not yet have repository-specific settings. If a repository has its own value, that repository setting takes priority. Put simply, organizations can now define operating rules such as “some reviews should go deeper, while others use the default.”
The engine behind it changed too
The more technical change GitHub announced alongside this is the way files are explored. Copilot code review now uses the grep, rg, glob, and view tools included in Copilot CLI and SDK in its code-navigation path. This replaces the previous separate custom file-navigation tool.
The goal of this change is to help Copilot find important files faster during review and reduce unnecessary exploration. GitHub explained that this adjustment, together with internal instruction tuning, makes reviews operate in a more focused way.
Why it matters for developers
AI code review is convenient, but when agent-style tools scan through code for a long time, cost and time increase as well. That is why the core of this announcement is not just a feature addition, but an operations update that makes review depth visible, manages defaults, and improves cost efficiency.
GitHub says that in offline and online evaluations, Copilot code review costs fell by about 20% while maintaining the existing review-quality bar. For teams, the next stage after “should we turn on AI review?” is deciding how much depth to apply to which kinds of reviews.
What not to misunderstand
This announcement is not a declaration that Copilot fully replaces human reviewers. The code review flow in the official documentation still works inside Pull Requests, where users check Copilot suggestions and humans review the results.
Medium analysis depth and organization defaults are also affected by public preview and settings availability. So rather than assuming every organization will immediately see the same screen and the same cost reduction, it is more accurate to treat this as an update for checking whether the setting is available in your own organization and how actual review costs change.
Sources and checked date · Published 2026-06-25 / Checked 2026-06-26T01:07:40+00:00
Sources
Leave a comment