🔍 Read the full analysis: GitHub’s Pull Request Review Workload Shifts As AI Agents Start More PRs on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
GitHub’s Octoverse 2026 report says more than 40% of pull requests on its platform began with a detectable coding agent in September, up from 17% in March. The company also introduced AI review and model-routing tools as review capacity becomes a bigger concern, but a merged pull request is not evidence that code is correct.
More than 40% of pull requests on GitHub in September began with a first draft from a coding agent the company could detect, up from 17% in March, according to GitHub’s Octoverse 2026 report, released October 28. The shift is increasing the volume of proposed code and putting more attention on review, approval and accountability—areas GitHub targeted with new Copilot tools announced this week.
GitHub reported 110 million agent-started pull requests in September, roughly six times the March volume. It defines an agent-started pull request as one whose first draft came from a coding agent it can detect. The company says the measurement is a floor because it cannot identify every use of an agent; it also counts a pull request even if a person later rewrites part of the draft. GitHub’s methodology includes public and private repositories unless a metric is explicitly limited, but the report does not give a public-versus-private breakdown for this figure.
GitHub’s data also show that 90% of detected agent pull requests were merged within two weeks in September, compared with 83% in February. For pull requests without a detected agent, the share stayed at about 86%. These figures measure whether changes were merged, not whether they were correct, safe or free of later rework. Agent-started pull requests also more often touched multiple files: 87% changed more than one file, compared with 79% of other pull requests, according to the report.
GitHub’s new offerings focus on the work after a draft is created. Copilot code review Max uses several AI reviewers from different model providers and combines their feedback. GitHub also made Copilot approvals eligible to count toward required approvals and released Copilot Fusion, a system that routes tasks to models for different steps. The company estimates an average review costs $0.25 for Lite, $1.50 for Balanced and $10 for Max, with GitHub Actions minutes charged separately.
Infographic 01 · The agent share
Four in ten pull requests now start with an agent.
GitHub’s own telemetry, March to September 2026. The first draft comes from an agent; review and merge stay with people.
Merged within two weeks
Share of pull requests merged within 14 days
Merged fast is not the same as correct. GitHub publishes no revert or defect rate.
Bigger changes to review
Share of pull requests that change more than one file
More files per change means more for a reviewer to check.
What GitHub counts: an “agent-started pull request” is one “whose first draft came from a coding agent that we can detect.” Undetected agents are missing, so GitHub calls the figure a floor. Metrics cover public and private repositories unless noted.
Source: GitHub Octoverse 2026, 28 Oct 2026 (text, chart captions, glossary and methodology). All figures are GitHub’s own.
Powered by Thorsten Meyer AI
Review Capacity Becomes the Constraint
The reported growth suggests that software teams may face a different constraint as code-generation agents become more common: reviewing and validating proposed changes, rather than producing an initial draft. A higher volume of pull requests can require more reviewer attention, particularly when changes span multiple files. GitHub’s new review products are a response to that pressure, while their listed prices make review a cost that teams may need to track alongside compute and other development infrastructure.
Approval rules also affect how teams distribute responsibility. GitHub says its cloud agent cannot approve or merge its own pull request, and the person who requested a change cannot approve it either. The company says developers remain accountable for code that ships. Those safeguards keep a human approval step in the workflow, but they do not by themselves establish that a review caught every defect. Organizations will need to decide where AI feedback or approval can count and how to measure quality through outcomes such as rework, reversions and incidents, not just pull-request totals.
For developers, the change may affect everyday work and training. Writing a clear task for an agent and critically reading its diff become more central skills. The source report also raises a workforce concern: small, straightforward coding tasks have often given junior developers practice, and those tasks are among the work agents can take on. The report does not establish how widely that is changing hiring or career paths.
Popular right now
Dyson V8 Cyclone Cordless Vacuum
Apple AirPods Pro 3 Wireless Earbuds
iRobot Roomba Max 715 Pet Robot Vacuum with AutoEmpty Dock
As an affiliate, we earn on qualifying purchases.
How GitHub Defines Agent Work
The figures come from GitHub’s annual Octoverse report, whose 2026 edition covers October 1, 2025, through September 30, 2026. GitHub uses the specific term “agent-started” in its definition: the first draft must come from an agent the company can detect. Elsewhere in the report, chart labels use terms such as “agent-generated” or “agent-authored.” For clarity, the reported share is best described using GitHub’s definition rather than treating every pull request with any AI assistance as agent-written.
The company recalculated its 2024 and 2025 figures using updated definitions and the report’s time window, so the revised historical data are not directly comparable with older Octoverse reports. At the wider platform level, GitHub reported 1.3 billion merged pull requests over the year, up from 538 million in 2025 and 402 million in 2024. Those totals show that the platform’s overall activity also grew; they do not isolate how much of that increase came from agents.
GitHub said agent activity rose around June, when agents began moving from single-session tools toward persistent cloud-based workflows that could track work over days. It also reported that 55% of Copilot monthly active users used agent features, up from 35% a year earlier, and that use of its cloud agent grew 230% year over year. These are company-reported platform metrics, not an independent measure of code quality or productivity.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in the Octoverse 2026 report
Quality Beyond the Merge Rate
GitHub’s report does not say how often merged agent-started changes later required rework, were reverted or caused incidents. Merging is a workflow outcome, not an independent test of correctness, security or maintainability. The report also does not provide a breakdown of the agent-started share by public and private repositories, or quantify how much agent use its detection misses.
The terminology varies across the report, and the figures depend on GitHub’s detection and definitions. Its reported 40%-plus share therefore should not be read as a measure of every pull request that received AI assistance. The company’s review-cost estimates are also its own estimates; the source does not detail the usage assumptions behind each tier or give a comparison with the labor cost of human review.
It is also unclear how many teams will allow Copilot approvals to satisfy required review rules, and what safeguards they will use when doing so. The report describes GitHub’s product capabilities and policies, but does not provide independent evidence that the new review tools reduce defects or reviewer workload.
Teams Will Test Review Policies
GitHub’s next practical test is whether its review and routing tools can keep pace with the increase in agent-started changes without adding unacceptable costs or weakening checks. Teams adopting them will need to set policies for which changes can receive AI review, when a human must approve, and how review credits and Actions minutes are budgeted.
As organizations put these tools into use, measures beyond volume will matter: reviewer time, defects found before release, rework, reverts and production incidents. GitHub’s report establishes a platform trend and announces tools intended to address it; it does not yet show how those tools perform across teams. Further evidence will be needed to assess whether they improve review capacity while preserving code quality and human accountability.
Key Questions
What does GitHub mean by an agent-started pull request?
GitHub counts a pull request when its first draft came from a coding agent the company can detect. The definition does not require the agent to have written the final version without human edits.
How many GitHub pull requests began with an agent in September?
GitHub reported 110 million agent-started pull requests in September 2026, representing more than 40% of pull requests by its measure. It said the share was 17% in March.
Does a high merge rate show that agent-written code is correct?
No. GitHub said 90% of detected agent pull requests were merged within two weeks, but merging does not establish correctness or show whether a change later required rework.
How much does GitHub estimate an AI code review costs?
GitHub estimates an average Copilot review at $0.25 for Lite, $1.50 for Balanced and $10 for Max. GitHub Actions minutes are additional.
Can GitHub’s cloud agent approve and merge its own pull request?
GitHub says its cloud agent cannot approve or merge its own pull requests. It also says the person who requested a change cannot approve it.
Source: ThorstenMeyerAI.com
