GitHub announced on June 19, 2026 that the Copilot usage metrics API now reports AI credits consumed per user. The new field is called ai_credits_used, and GitHub says it appears in both the single-day users-1-day and 28-day users-28-day user-level reports at the enterprise and organization levels.
This is a small changelog item, but it matters for QA teams using AI coding assistants for test automation, flaky-test triage, pull request review, and agent-style repo tasks. Until now, many teams could see adoption signals, but not a clean per-user indicator of how much Copilot credit consumption each engineer actually drove.
What GitHub changed on June 19
- GitHub added a new
ai_credits_usedfield to user-level Copilot usage reports. - The field is available in both the
users-1-dayandusers-28-dayreports. - GitHub says the value is derived from the same AI credits consumption data used in the usage-based billing API.
- The metric is available at the enterprise and organization levels for admins who already have access to Copilot usage metrics through the REST API.
GitHub also makes one limit explicit: ai_credits_used is an overall per-user total. It is not currently broken down by feature, model, or surface. That means you can see who is consuming credits, but not yet whether those credits came from chat, code review, agent tasks, or another Copilot workflow.
Why this matters for QA engineers
GitHub Copilot AI credits per user is useful because QA automation work is often harder to measure than ordinary coding output. A tester might use Copilot to refactor Playwright fixtures, review brittle selectors, summarize CI failures, or draft negative-path API coverage. Those workflows create value, but they can also create cost. With per-user credit reporting, QA leaders have a better chance to connect usage with actual outcomes.
- Better pilot measurement: compare Copilot-heavy QA users with changes in flaky-test backlog, PR turnaround, or automation maintenance time.
- Cleaner governance: identify teams adopting Copilot aggressively before budgets or review standards drift.
- More honest rollout conversations: usage can now be discussed with data, not just anecdotes about who “uses AI a lot.”
A practical QA use case
Imagine a QA platform team rolling out Copilot to a group that maintains Selenium and Playwright suites across several repositories. One group uses Copilot mainly for low-risk cleanup and debugging help. Another group starts relying on agent-heavy workflows for code review, test repair, and CI edits. If both groups say adoption is going well, the next question is obvious: how much usage are they actually driving, and what result is that usage buying?
The June 19 metrics update does not answer everything, but it gives teams one missing piece. You can now compare per-user credit consumption against QA outcomes you already track, such as:
- flaky-test reopen rate
- time to stabilize failing CI checks
- pull request review turnaround for automation changes
- number of test-maintenance commits merged safely
- escaped defects linked to automation gaps
If a high-usage pattern correlates with better maintenance speed and no drop in test quality, that is a stronger adoption signal than simple seat activation. If usage climbs without any improvement in test stability or review quality, the team has evidence to tighten prompts, review rules, or tool scope.
Starter API request
curl -L \
-H "Accept: application/vnd.github+json" \
-H "Authorization: Bearer <TOKEN>" \
-H "X-GitHub-Api-Version: 2026-03-10" \
"https://api.github.com/orgs/ORG/copilot/metrics/reports/users-28-day/latest"That request returns report download links for the current 28-day organization users usage report. From there, inspect the user-level report for the new ai_credits_used field.
What this update does not solve yet
- It does not break credit usage down by feature, model, or surface.
- It is a metrics signal, not a billed total, so invoicing still belongs to billing data.
- It does not tell you whether the credits produced good QA outcomes. Teams still need their own quality and productivity baselines.
That last point matters most. QA organizations should avoid treating higher AI usage as automatic progress. Good governance means pairing cost signals with evidence such as assertion quality, flakiness trend, reviewer effort, and production defect escape rate.
Three checks QA teams should run this week
- Identify the QA users or repos in your Copilot pilot and compare their credit usage with current test-maintenance or triage outcomes.
- Set a review rule for high-usage workflows so heavy agent or code-review usage still gets human validation on selectors, waits, assertions, and CI changes.
- Decide which metric will prove value before expanding Copilot access, such as flaky-test reduction or faster repair of broken automation.
Conclusion
The June 19, 2026 GitHub update is not flashy, but GitHub Copilot AI credits per user is the kind of operational detail mature QA teams need. It gives engineering and test leaders a clearer way to measure whether AI coding-assistant usage is producing enough value to justify its cost. For automation testers, that means better conversations about rollout, governance, and where Copilot is actually helping the work.