🔍 Read the full analysis: A New GitHub Pull Request Pattern: AI Starts 4 In 10, Review Follows on ThorstenMeyerAI.com
Get hardware and tech essentials 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, up from 17% in March. The company also introduced tools for AI-assisted review, with estimated per-review costs, while keeping responsibility for shipped code with people.
Coding agents started more than 40% of pull requests on GitHub in September, up from 17% in March, according to the company’s 2026 Octoverse report, released October 28 at its Universe conference. GitHub says the September total was 110 million agent-started pull requests, a shift that puts more pressure on developers to review and approve code drafted by software agents.
GitHub defines an agent-started pull request as one whose first draft came from a coding agent the company can detect. The company says its count is a floor because some agent use may not be detected. A pull request is a proposed change to a codebase; it typically goes through review before being merged. The report’s metric describes how a change began, not how much of the final code was written or revised by people.
In September, 90% of detected agent pull requests were merged within two weeks, compared with 83% in February. Pull requests without a detected agent had a merge rate of about 86%, GitHub reported. Those figures measure whether changes were merged, not whether they were correct, safe or free of later rework.
GitHub announced several review-related products and changes at its conference. Copilot code review Max uses reviewers from multiple model providers and combines their feedback; Copilot approvals can count toward required approvals; and Copilot Fusion routes different tasks to models. GitHub estimates an average review costs $0.25 in Lite mode, $1.50 in Balanced mode and $10 in Max mode, 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 Moves to Center Stage
The figures point to a change in where software teams may spend time. If agents draft more proposed changes, review capacity, approval policies and testing become more prominent constraints. A higher volume of pull requests can mean more work for maintainers, even when each individual draft is quick to produce. GitHub’s new review tools are aimed at that pressure, but their cost and effectiveness will depend on how teams use them.
The distinction between merged and correct matters. A merge-rate figure alone does not show whether a change caused defects, needed extensive rework or was later reverted. Companies assessing agent use would need to track those outcomes alongside output volume and review time. The reported per-review prices also make AI review a budget question, particularly for organizations that use the more expensive multi-reviewer mode at scale.
There are implications for responsibility and developer work. GitHub says people remain accountable for code that ships, and its cloud agent cannot approve or merge its own pull requests. The source report also says the person who requested a change cannot approve it. These controls preserve a human sign-off step, but they do not by themselves establish how thoroughly a change has been examined. Developers still need to understand the code and judge whether it meets the task and quality requirements.
Products Worth Considering
Logitech MX Mechanical Full-Size Wireless Mechanical Keyboard, Tactile Quiet Switches, Backlit Keys, Bluetooth, USB-C, Multi-Device, Graphit
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
Logitech MX Mechanical Wireless Illuminated Keyboard Tactile (Graphite)
As an affiliate, we earn on qualifying purchases.
How GitHub Counts Agent Drafts
The Octoverse is GitHub’s annual report on activity across its platform. The 2026 edition covers October 1, 2025, through September 30, 2026. GitHub says its metrics cover public and private repositories unless a metric is specifically marked otherwise; the report does not provide a public-versus-private split for the agent-started pull-request figure.
The report uses several nearby terms, including “agent-started,” “agent-generated” and “agent-authored.” Its formal definition centers on the first draft coming from a detectable coding agent, so “agent-started” is the most precise description of what the headline percentage measures. The count does not mean that every line in those pull requests was written by an agent.
GitHub also recalculated its 2024 and 2025 figures using updated definitions and a revised reporting window, so earlier Octoverse totals are not directly comparable without that qualification. Across the platform, the report says merged pull requests reached 1.3 billion in the 12 months to September, compared with 538 million in 2025 and 402 million in 2024. GitHub attributes a broader surge in activity partly to agents becoming more persistent and able to work in the cloud.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in its Octoverse report
Limits of the September Measure
The report does not disclose how many agent-started pull requests were in public repositories versus private company repositories, or how GitHub validates the quality of the changes after merge. Because the company counts only agents it can detect, the actual share of agent-started work is unknown. The definition also allows a pull request to count as agent-started even if a person substantially revises the draft afterward.
The 90% merge figure does not establish that the code was correct or that it remained in production without incident. The source material does not give comparable figures for post-merge defects, reversions, review time or maintenance costs. It is also too early to know how often teams will use Copilot’s approval feature as a required approval or whether the new review products will reduce reviewer workload in practice.
Teams Put Review Rules to Work
GitHub’s newly announced review features give teams options for routing model-based feedback and deciding whether Copilot approvals count toward repository requirements. The next evidence will come from how organizations configure those controls, how much they spend on review credits and Actions minutes, and whether their workflows retain meaningful human scrutiny.
For engineering teams, useful measures include review time, follow-up changes, reverts and incidents, rather than pull-request volume alone. GitHub’s next reporting and product updates may clarify adoption and outcomes, but the current report does not set out a timetable for those data. Until then, the central confirmed development is a sharp rise in the share of pull requests that GitHub can identify as agent-started, not proof that those changes are better or worse than human-started work.
Key Questions
What does GitHub mean by an agent-started pull request?
It means the first draft came from a coding agent GitHub can detect. The label does not mean the agent wrote every part of the final change.
How many GitHub pull requests were agent-started in September?
GitHub reported 110 million, or more than 40% of pull requests that month. The company says the figure is a floor because not all agent use is detectable.
Does a high merge rate show that agent code is correct?
No. GitHub said 90% of detected agent pull requests were merged within two weeks, but merging is not a measure of correctness. The report does not provide corresponding figures for defects, reversions or later rework.
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, plus GitHub Actions minutes. Actual spending depends on usage and configuration.
Can a GitHub coding agent approve or merge its own pull request?
According to the source report, GitHub’s cloud agent cannot approve or merge its own pull request. The report also says the person who requested the change cannot approve it.
Source: ThorstenMeyerAI.com
