🔍 Read the full analysis: From AI-Generated Pull Requests To Human Review: GitHub’s Changing Workflow on ThorstenMeyerAI.com
Get audio and creator gear 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 in September 2026 began with a detectable coding agent, or 110 million proposals. GitHub’s new tools focus on reviewing and routing AI-generated changes, but the company says people remain accountable for code that ships; the reported figures do not establish that merged code is correct.
More than 40% of pull requests on GitHub in September started with a coding agent’s first draft, the company said in its Octoverse report released October 28. That amounted to 110 million agent-started pull requests, up from 17% in March, as GitHub introduced tools aimed at reviewing and routing AI-generated changes rather than only producing more code.
GitHub defines an agent-started pull request as one whose first draft came from a coding agent the company can detect. Its estimate covers public and private repositories unless otherwise specified, but GitHub did not provide a public-versus-private breakdown for this measure. The company describes the figure as a floor because some agent use may go undetected. A human can revise a draft substantially and the pull request still counts as agent-started.
The report says 90% of agent pull requests were merged within two weeks in September, compared with 83% in February. Pull requests without a detected agent remained at about 86%. Those are merge rates, not measures of correctness, code quality or the number of changes later reverted. GitHub also reports that 87% of agent-started pull requests changed more than one file, against 79% of other pull requests, potentially giving reviewers more material to examine.
At its Universe conference, GitHub launched Copilot code review Max, which uses several AI reviewers from different model providers and combines their reviews. The company also made Copilot approvals eligible to count toward required approvals and released Copilot Fusion, a system that routes tasks to models. GitHub estimates average Copilot review costs of $0.25 for Lite, $1.50 for Balanced and $10 for Max, plus GitHub Actions minutes. The report does not say how these tools will affect review quality or total engineering costs.
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 New Bottleneck
The figures point to a shift in where development work may take time: if agents draft more changes, people still need to decide whether those changes are safe and appropriate to merge. The increase matters because pull requests are a central checkpoint before code enters a project, and the reported rise in multi-file changes may increase the scope of review. But the adoption numbers alone do not show that teams are shipping better software or saving time.
GitHub’s pricing estimates also make review a budgeting decision as well as a workflow choice. Companies may need to account for review credits, compute and Actions usage, while deciding when an AI review can count toward a required approval. The reported merge rate should not be treated as evidence of correctness: a merged change can still contain defects, require rework or later be reverted. Teams need outcome measures such as incidents, rework and reversions, not only counts of generated or merged pull requests.
For developers, the change may shift effort from writing a first draft toward defining tasks precisely and checking diffs carefully. The source argues that small starter tasks, often used by junior developers to learn, are among the kinds of work agents can take on. That is a concern about how training and entry-level work may change, not a finding established by GitHub’s adoption statistics.
Products Worth Considering
Logitech MX Mechanical Wireless Illuminated Keyboard Tactile (Graphite)
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How GitHub Counts Agent Work
The Octoverse is GitHub’s annual report on activity across its platform. Its 2026 edition covers October 1, 2025, through September 30, 2026. GitHub says the agent share is based on signals it can detect, including agent-opened pull requests. The report uses several nearby terms—such as “agent-started,” “agent-generated” and “agent-authored”—but the stated definition is specifically about who produced the first draft.
GitHub says it recalculated 2024 and 2025 figures using revised definitions and time windows, so those earlier Octoverse reports are not directly comparable. The company also reported broader growth in platform activity: 1.3 billion pull requests were merged during the year, compared with 402 million in 2024 and 538 million in 2025. Those totals describe overall platform activity and should not be read as agent-only counts.
GitHub’s report says Copilot cloud-agent use grew 230% year over year and agent chat use grew 85%. It also says 55% of Copilot monthly active users used agent features, up from 35% a year earlier. These are company-reported platform metrics; the source material does not provide independent verification or enough methodological detail to assess each measure beyond GitHub’s definitions.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in its Octoverse 2026 report
What the Adoption Figures Cannot Show
The report does not establish whether merged agent-started changes are correct, how often they cause incidents, or how much human review and rework they require. GitHub reports merge rates but does not provide a quality comparison in the supplied material. It also does not disclose a split of the 110 million pull requests between public and private repositories, or the number missed because agents were not detectable.
There is also no detailed breakdown of how many pull requests used each review product, how often AI recommendations changed a reviewer’s decision, or how the estimated per-review prices translate into total costs for a team. GitHub says its cloud agent cannot approve or merge its own pull requests, and that the person who requested a change cannot approve it either. The source does not fully describe how approval rules work across every organization or repository configuration.
Finally, the source material includes a customer example and industry figures, but those should not be generalized to all companies. GitHub’s reported usage metrics show adoption on its platform; they do not by themselves establish the effect on productivity, software quality or developer employment.
How Teams May Adapt Review Rules
GitHub has released the review and routing features, but the supplied material does not name a future product milestone or publication date for further results. The next evidence to watch is how organizations use the tools: whether they allow Copilot approvals to satisfy required review counts, how they set limits on review tiers, and how they monitor changes that reach production.
For teams adopting agent workflows, practical next steps include setting clear approval policies, budgeting for review usage and Actions minutes, and tracking rework, reversions and incidents alongside pull-request volume. Developers and managers will also need to establish how much human inspection is expected before a change is merged. GitHub’s September figures show the scale of detected agent activity; the operational and quality consequences remain to be measured.
Key Questions
How many GitHub pull requests started with an AI agent in September?
GitHub reported 110 million agent-started pull requests in September 2026, representing more than 40% of pull requests on the platform that month.
What does GitHub mean by an agent-started pull request?
It is a pull request whose first draft came from a coding agent GitHub can detect. GitHub says its measure is a floor because some agent use may not be detected.
Does a high merge rate mean AI-generated code is correct?
No. GitHub’s reported merge rate measures whether pull requests were merged within a period. It does not establish that the code was correct, defect-free or free of later rework.
What new GitHub review tools were announced?
GitHub announced Copilot code review Max, which combines reviews from multiple AI reviewers, made Copilot approvals eligible to count toward required approvals, and released Copilot Fusion for routing tasks to models.
Who is responsible for AI-assisted code that ships?
GitHub says developers remain accountable for shipped code. 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
