🔍 Read the full analysis: What Comes After AI Starts 4 In 10 GitHub Pull Requests? The Review on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
GitHub’s Octoverse report says more than 40% of pull requests on its platform started with a detectable coding agent in September 2026, up from 17% in March. The company introduced AI review and model-routing tools as agent-written changes become more common, but a merged pull request is not proof that code is correct, and GitHub says developers remain accountable for what ships.
More than 40% of pull requests on GitHub in September 2026 started with a coding agent’s first draft, up from 17% in March, according to the company’s Octoverse report released October 28. GitHub says the September total was 110 million agent-started pull requests, a shift that puts greater emphasis on reviewing and approving code, not just producing it.
GitHub defines an agent-started pull request as one whose first draft came from a coding agent the platform can detect. The company describes the count as a floor, because it cannot identify every instance of agent use. The figure includes public and private repositories under the report’s general methodology, but GitHub does not provide a separate breakdown for the two.
GitHub also reported that 90% of agent pull requests were merged within two weeks in September, compared with 83% in February. Pull requests without a detected agent remained near 86%. Those rates measure whether changes were merged, not whether they were correct, safe or free of later rework.
At its Universe conference, GitHub announced Copilot code review Max, which draws on several AI reviewers from different model providers and combines their feedback; changes that let Copilot approvals count toward required approvals; and Copilot Fusion, a system that routes tasks to models. GitHub estimates an average Copilot review costs $0.25 in Lite, $1.50 in Balanced and $10 in Max, in addition to GitHub Actions minutes.
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 Becomes the Capacity Constraint
The report points to a change in where software teams may spend their time: if agents can draft changes quickly, human review and approval can become the limiting step. A higher volume of pull requests does not automatically mean faster or better software. Reviewers still need to understand what a change does, whether it fits the system and what could break when it is deployed.
GitHub says 87% of agent-started pull requests changed more than one file, compared with 79% of pull requests without a detected agent. More files can mean a wider review surface, although the statistic alone does not establish that these changes are harder to assess or more error-prone.
For companies, the practical issue is how to allocate reviewer time and pay for AI review alongside compute and workflow costs. Teams will need to judge the tools by outcomes such as rework, reversions and production incidents, rather than counting drafts or merges alone. For developers, writing clear tasks and critically checking diffs remain central skills. The source report argues that small tasks often used to train junior developers are also among the work agents can take on; how that affects hiring and training is not established by the figures.
Products Worth Considering
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming Monitor
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How GitHub Counts Agent Drafts
The Octoverse report covers October 1, 2025, through September 30, 2026 and presents GitHub’s own platform data. The 40%-plus figure refers to pull requests that began with an agent’s first draft, not every pull request that received any AI assistance. A human may substantially revise a draft and it still qualifies under that definition.
The report uses several related descriptions, including “agent-started,” “agent-generated” and “agent-authored.” For precision, the reported measure is best described as agent-started. GitHub also recalculated 2024 and 2025 figures using updated definitions and the report’s measurement window, so comparisons with older Octoverse reports may not be direct.
The rise occurred amid broader platform growth. GitHub reported 1.3 billion merged pull requests during the year, compared with 538 million in 2025 and 402 million in 2024. It also said 55% of Copilot monthly active users used agent features, up from 35% a year earlier. These figures describe GitHub’s platform and do not by themselves show how much of the growth was caused by agents.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub’s Octoverse report
What the Adoption Figure Cannot Show
GitHub has not published a split of the agent-started figure between public and private repositories, nor a measure of how often an agent’s draft was substantially rewritten by a person. Because the platform counts only agents it can detect, the actual share is unknown. The reported change in wording across parts of the report also makes it important to retain the company’s precise “started with a first draft” definition when citing the statistic.
The merge rate does not establish code quality. The supplied figures do not show whether agent-started changes led to more defects, security problems, reversions or maintenance work, or how results vary by project and team. It is also unclear how often teams will rely on AI review in approval processes or whether the new review tiers will reduce human workload in practice.
The source material includes an Accenture manager’s claim that AI authors 91% of his team’s commits and that production errors fell, but it provides no full methodology or comparison details for that claim. It should not be treated as evidence that agent-written code generally reduces errors.
Teams Must Measure Review Outcomes
GitHub’s newly announced review and routing tools put more review options in front of teams, with listed per-review estimates ranging from $0.25 to $10 plus Actions minutes. The material does not give a rollout timetable or explain all plan-specific access terms. Organizations will need to decide which changes, if any, can receive an AI approval that counts toward required approvals, while accounting for GitHub’s stated limits on its own cloud agent.
GitHub says its cloud agent cannot approve or merge its own pull requests, and the person who requested a change cannot approve it either. Those rules keep approval separate from the agent that produced the work, but they do not answer how companies will set review policies across different risk levels.
The next useful evidence will be how teams use these tools and whether they change review time, rework, reversions and incident rates. Until those outcomes are reported with clear methods and comparisons, the 40%-plus figure is evidence of adoption on GitHub—not a measure of productivity or software quality.
Key Questions
What does GitHub mean by an agent-started pull request?
It means the pull request’s first draft came from a coding agent GitHub can detect. A person can later revise the draft and the pull request still counts.
Does the 40% figure cover every use of AI to write code?
No. GitHub says it counts detectable agent signals and calls the result a floor because some agent use may not be detected. The figure is not a count of every instance of AI assistance.
Does a high merge rate mean agent-written code is correct?
No. GitHub reported that 90% of agent pull requests were merged within two weeks, but merging is not a code-quality measure. The figures do not establish whether changes caused defects or required later rework.
How much does Copilot code review cost?
GitHub estimates an average review at $0.25 for Lite, $1.50 for Balanced and $10 for Max, plus GitHub Actions minutes. The source material does not specify all plan or usage conditions.
Who is responsible for code that an agent helps write?
GitHub says developers remain accountable for what ships. Its cloud agent cannot approve or merge its own pull request, and the person who requested a change cannot approve it either.
Source: ThorstenMeyerAI.com
