🔍 Read the full analysis: AI Agents Are Starting GitHub Pull Requests; How Will Teams Review Them? 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’s Octoverse report says more than 40% of pull requests in September 2026 began with a coding agent’s first draft, compared with 17% in March. The company introduced AI review and model-routing tools, but agent detection is incomplete, and merged pull requests are not necessarily correct. GitHub says developers remain accountable for code that ships.
More than 40% of pull requests on GitHub in September 2026 began with a coding agent’s first draft, up from 17% in March, according to the company’s new Octoverse report. Released October 28 at GitHub Universe, the figures show how quickly agents are entering software teams’ code workflows—and put more pressure on how developers review, approve and take responsibility for changes.
GitHub counted 110 million agent-started pull requests in September, about six times the March volume. The company defines an agent-started pull request as one whose first draft came from a coding agent it can detect. It says the count is a floor rather than a complete tally, because some agent use may go undetected. The report’s methodology includes public and private repositories unless a metric is specifically excluded; GitHub did not provide a public-versus-private split for this figure.
The report also says 90% of agent pull requests were merged within two weeks in September, compared with 83% in February. The rate for pull requests without a detected agent was about 86%. A merge records that a change was accepted into a codebase; it does not establish that the change was correct, safe or free of later rework.
GitHub announced Copilot code review Max, which brings together reviews from several AI models, and Copilot Fusion, a system that routes tasks to models. It also made Copilot review approvals eligible to count toward required approvals. GitHub 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 Work Is Moving to Center Stage
The numbers point to a shift in where software teams may spend time. When agents produce a first draft quickly, reading and testing the change can become a larger share of the work. GitHub’s report says 87% of agent pull requests changed more than one file, compared with 79% of pull requests without a detected agent. Changes spanning more files can require reviewers to track more interactions across a codebase; the statistic alone does not say whether those changes were harder or less reliable.
Organizations may need to plan for review capacity and review-tool costs, not just the cost of generating code. They will also need clear rules for where AI-generated reviews can count toward approval requirements. Teams can track pull-request volume, but measures such as rework, reversions and production incidents are needed to understand outcomes. GitHub’s figures describe activity and merging, not the quality of every submitted change.
The change also affects developers’ work and training. Writing clear task instructions and critically checking diffs become central skills when an agent does more of the initial drafting. The source report argues that small tasks once used to build junior developers’ experience may be among the first work agents take on. How teams replace that learning opportunity is an open workforce question, not something the adoption figures answer.
Products Worth Considering
HP Sprocket Photo Booth Machine Instant Color Photo Printer with 10.1 Touchscreen, LED Ring Light, 250+ Frames, 3×4 ZINK Paper, QR Code Shar
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming Monitor
FLEXISPOT EN1-PRO Electric Standing Desk, 79 x 32, Black Walnut
As an affiliate, we earn on qualifying purchases.
How GitHub Defines Agent-Started Work
The statistics come from GitHub’s Octoverse 2026, its annual review of activity on the platform. This edition covers October 1, 2025, through September 30, 2026. GitHub’s definition is based on whether the first draft came from an agent it can detect; a human can revise the draft substantially and the pull request still counts. That makes the figure a measure of how work began, not how much code an agent ultimately wrote.
The report uses several related labels, including “agent-started,” “agent-generated” and “agent-opened.” For the headline share, “agent-started” is the more precise wording in the stated definition. GitHub also recalculated its 2024 and 2025 figures under new definitions and a new reporting window, so those historical totals are not directly comparable with figures from earlier Octoverse reports.
GitHub describes broader platform growth alongside the agent data: it reports 1.3 billion merged pull requests in the 12 months to September, up from 538 million in 2025. Those totals reflect overall activity, not agent work alone. The company’s account of a surge in activity around June links it to agents moving from single-session tools toward persistent cloud workflows, but the report does not establish that agents alone caused the platform-wide increase.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in its Octoverse report
What the Adoption Figures Cannot Show
The actual share of agent use may be higher than 40%, because GitHub’s detection does not capture every agent. The report does not disclose how many uses go undetected or provide a split between public and private repositories for the agent-started figure. It also does not show how much of each draft was later rewritten by a person.
The figures do not establish whether agent-written changes have higher or lower defect rates, or whether faster merging leads to better software. Merging is not a quality measure, and the report’s merge-rate comparison does not account for later reverts, incidents or maintenance costs. GitHub’s review price estimates likewise do not give each organization’s final bill, which can depend on usage and additional Actions minutes.
It remains unclear how companies will set approval policies for AI reviews, and how junior developers will gain experience if agents take on routine starter tasks. GitHub’s announced tools describe available product options; they do not show how widely teams will adopt them or whether they will reduce the human review workload.
Teams Must Set Review Rules
The immediate next step for engineering teams is to decide which changes require human approval, whether an AI review can count toward required approvals, and how to test agent-authored work before it ships. Teams using GitHub’s review products can also monitor usage against the listed price tiers and account for GitHub Actions minutes. The company has not provided, in the supplied report, a timetable for broader rollout or adoption of the new review options.
For evaluating results, teams can compare rework, reversions and production incidents alongside merge rates and pull-request volume. Those measures would help distinguish faster throughput from reliable software delivery. GitHub says developers remain responsible for shipped code, and its stated restrictions on cloud-agent approvals leave a human approval step in place for the workflows described.
More evidence should come from subsequent platform reporting and from organizations’ own operating data. Until then, GitHub’s September figures establish a rapid rise in detected agent-started work, but not the overall quality, cost-effectiveness or long-term effect of that work on software teams.
Key Questions
What does GitHub mean by an agent-started pull request?
GitHub says it is a pull request whose first draft came from a coding agent the company can detect. A person can later revise the draft and it still counts.
Did agents start more than 40% of all pull requests?
GitHub reports that more than 40% of pull requests in September began with a detected agent draft. It calls the count a floor because not all agent use may be detected, so the figure is not a complete tally.
Does a high merge rate show that agent-written code is correct?
No. GitHub says 90% of agent pull requests were merged within two weeks in September, but a merge does not prove correctness or safety. The report does not provide defect or rework rates for the comparison.
How much does GitHub estimate AI code review costs?
GitHub estimates an average review at $0.25 for Lite, $1.50 for Balanced and $10 for Max. GitHub Actions minutes are additional, and actual costs depend on usage.
Who is responsible for code an agent submits?
GitHub says developers remain accountable for what ships. Its stated rules bar the cloud agent from approving or merging its own pull requests and bar the requester from approving the change.
Source: ThorstenMeyerAI.com
