🔍 Read the full analysis: 4 In 10 GitHub Pull Requests Are AI-Started—Reviewers Face A Shift on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
GitHub says more than 40% of pull requests on its platform in September 2026 began with a detectable coding agent, up from 17% in March. The company introduced new review and model-routing tools as agent-generated changes increase, but its counts do not show whether merged code was correct or how agent use divides between public and private repositories.
More than 40% of GitHub pull requests in September began with a first draft from a detectable coding agent, up from 17% in March, according to the platform’s Octoverse 2026 report. GitHub said the September figure represented 110 million agent-started pull requests, a shift that has put greater attention on how quickly developers can inspect, test and approve proposed changes.
GitHub released the report on October 28 at its Universe conference. It defines an agent-started pull request as one whose first draft came from a coding agent the company can detect. GitHub describes the measurement as a floor because some agent use may not be detected. The figure includes public and private repositories, according to the report’s general methodology, but GitHub did not publish a breakdown between them.
The report says 90% of agent-started pull requests were merged within two weeks in September, compared with 83% in February. Pull requests without a detected agent held at about 86%, according to GitHub. These are merge rates, not measures of correctness: the report does not establish that merged changes were bug-free, safe or beneficial, nor does it show how often they later needed rework.
GitHub also announced features focused on review and workflow. Copilot code review Max combines reviews from multiple model providers into one response; Copilot approvals can count toward required approvals; and Copilot Fusion routes different steps to selected models. GitHub estimates an average review will cost $0.25 for Lite, $1.50 for Balanced and $10 for Max, plus GitHub Actions minutes. Those are company estimates, not a full accounting of review costs for every team.
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 Work Becomes the Constraint
The figures suggest that AI is changing where time is spent in software development: agents can produce an initial change quickly, but a person still has to decide whether it is fit to ship. If proposed changes arrive faster than teams can inspect them, review capacity, testing and approval practices may become limiting factors. GitHub’s new review products address that pressure, while also adding per-review charges and potential Actions usage that organizations will need to account for.
The change also affects how engineering teams measure output. A larger number of pull requests or a higher merge rate alone cannot show whether software quality improved. Teams will need to track outcomes such as rework, reversions and production incidents alongside volume, and decide where AI-generated review can count toward formal approval. For developers, the work increasingly includes writing precise task instructions and critically examining a diff rather than treating an agent’s draft as ready to ship.
Popular right now
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Ninja Foodi 10-Quart DualZone 6-in-1 Air Fryer, Black, DZ302
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
As an affiliate, we earn on qualifying purchases.
How GitHub Measures Agent Work
The Octoverse report covers October 1, 2025, through September 30, 2026 and presents GitHub’s annual view of activity on its platform. GitHub says developers made 17.9 billion pushes and 34.3 billion commits during the period. Merged pull requests reached 1.3 billion, compared with 538 million in 2025 and 402 million in 2024, based on the report’s figures. GitHub recalculated its 2024 and 2025 numbers under updated definitions and time windows, so earlier reports are not directly comparable.
The report’s wording varies between “agent-started,” “agent-generated” and “agent-authored.” Its definition centers on the first draft and detectable agent signals. A pull request can therefore remain counted as agent-started even if a person later substantially changes it. GitHub also reports that 87% of agent-started pull requests changed more than one file, compared with 79% of pull requests without a detected agent; that indicates broader diffs, not necessarily lower quality.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in its Octoverse 2026 report
What the Report Cannot Show
The report does not reveal the public-versus-private repository split for agent-started pull requests, or how much of the increase came from particular industries, companies or coding agents. Because GitHub’s method counts only agents it can detect, the actual share may differ from the reported figure. The company’s wording also varies across the report, though its definition refers specifically to a detectable agent’s first draft.
Most importantly, a pull request being merged does not establish that its code was correct or that it performed well in production. GitHub’s reported merge rates do not provide a comparative account of defects, security issues, review time or later rework. Nor does the supplied report explain how teams should evaluate an AI review that counts toward a required approval beyond the announced product controls.
Teams Face Review and Cost Decisions
GitHub’s announced Copilot review and routing features put the next test in day-to-day use: whether they help teams assess a growing stream of changes without weakening review standards. Organizations will need to determine which changes can use AI review, which require a human approval, and how to budget for review tiers and Actions minutes. GitHub says its cloud agent cannot approve or merge its own pull requests, and the person who requested a change cannot approve it either.
Further reporting will be needed to establish whether the new tools reduce review delays or improve code outcomes. Useful measures would include time to review, defects, reversions, incidents and the amount of human rework, rather than pull request counts alone. Until those results are available, the report documents rapid adoption and a shift in workflow, but not proof that agent-started code is better or worse.
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 it can detect. Human edits made afterward do not necessarily change that classification.
Does the 40% figure cover every AI-assisted pull request?
No. GitHub says some agent use may not be detected and calls the figure a floor. It has not published a separate count for undetected agent assistance.
Does a high merge rate mean agent-written code is reliable?
No. GitHub’s reported merge rate shows how many pull requests were merged within a stated period. It does not establish whether the code was correct, secure or free from later rework.
How much does GitHub estimate Copilot code review will cost?
GitHub estimates an average review at $0.25 for Lite, $1.50 for Balanced and $10 for Max, plus GitHub Actions minutes. Actual costs may depend on use and configuration.
Who is responsible for code that an agent helps produce?
GitHub says developers remain accountable for what ships. The company also says its cloud agent cannot approve or merge its own pull requests, and the person who requested a change cannot approve it.
Source: ThorstenMeyerAI.com
