🔍 Read the full analysis: How The Growth Of AI-Started GitHub PRs Is Changing Code 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 in September 2026 began with a detectable coding agent, up from 17% in March. The growth is shifting attention from producing code to reviewing it: GitHub announced new AI review and model-routing tools, while saying 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 Octoverse report released October 28. The shift is putting code review, approval and the cost of checking changes under greater pressure, and GitHub announced new Copilot tools aimed at those tasks.
GitHub says the September share represents 110 million agent-started pull requests in one month, around 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 because some agent use may go undetected. The measure includes public and private repositories; GitHub did not provide a split 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%. A merge records that a change entered a codebase; it does not establish that the change was correct or free of problems.
GitHub’s announcements focus in part on review. Copilot code review Max combines reviews from several AI model providers into one result, and GitHub said Copilot approvals can count toward required approvals. Copilot Fusion routes different steps of a task to models it selects. 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 Becomes the New Constraint
When agents can produce a first draft quickly, the work of deciding whether that draft is safe and useful can take a larger share of a team’s time. GitHub’s figures point to a substantial increase in proposed changes, but they do not show how much reviewer time each change takes or whether the additional code improves software.
The change also has budget and workflow implications. Teams may need to weigh review-tool charges and Actions usage against reviewer capacity, and decide whether an AI-generated approval should count toward a repository’s rules. For developers, the source material’s practical implication is that writing a clear task and scrutinizing a diff remain important even when an agent supplies much of the code.
Hiring and training are also part of the concern. Small, bounded changes have often given less-experienced developers practice reading and modifying code. If agents take on more of those tasks, teams may need to make deliberate room for junior developers to build those skills. The report does not measure whether that is already happening.
Products Worth Considering
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How GitHub Defines Agent Work
The Octoverse report covers October 1, 2025, through September 30, 2026, and uses GitHub’s platform data. Its precise measure is “agent-started”: a detectable coding agent supplied the first draft. A human may still substantially revise that draft, so the statistic does not mean an agent wrote every line ultimately merged.
GitHub recalculated its 2024 and 2025 figures using updated definitions and the report’s time window, meaning earlier Octoverse totals are not directly comparable. The report also says merged pull requests rose to 1.3 billion over the year, from 538 million in 2025, alongside broader growth in platform activity. Those platform-wide totals provide scale, but do not establish that agents alone caused the increase.
GitHub says developers remain accountable for code that ships. Its cloud agent cannot approve or merge its own pull requests, and the person who requested a change cannot approve it either. Those safeguards place a human in the approval process, though they do not by themselves show how effectively a review catches defects.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in its Octoverse report definition
What the Metrics Do Not Show
GitHub has not published an open-source versus private-repository breakdown for the agent-started share, nor a detailed account of which coding agents its detection captures. Since undetected use is excluded, the reported share is not a complete census. The report’s wording also varies in places, using terms such as “agent-generated” and “agent-authored”; “agent-started” is the more precise description of the stated measure.
The merger rate does not reveal whether changes were correct, later reverted, or associated with production incidents. The cited figures also do not quantify review time, the share of approvals made by AI, or the effect of the new tools on defect rates. GitHub’s cost estimates are estimates per review, and total spending will depend on usage and additional Actions minutes.
Teams Weigh Review Rules and Cost
GitHub’s report and product announcements put implementation choices in front of engineering teams: whether to use AI review, which review tier fits a repository, and whether Copilot approvals count toward required approvals. The source material does not give a schedule for adoption or independent results showing how the tools perform in production.
As agent use grows, a clearer picture will depend on measures beyond pull-request counts: reviewer workload, rework, reversions and incidents. GitHub’s next reports or product data may show whether faster draft production is matched by dependable review. For now, the reported growth is clear; its effect on code quality and team capacity remains unsettled.
Key Questions
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. A person may later edit the draft. GitHub says undetected agent use is missing from the count.
How many pull requests did agents start in September?
GitHub reported 110 million in September 2026, or more than 40% of its pull requests that month. It said the share was 17% in March.
Does a high merge rate mean agent code is correct?
No. GitHub reported that 90% of agent-started pull requests were merged within two weeks, but merging is not proof of correctness. The report does not provide a corresponding measure of defects or later reversions.
How much does GitHub estimate AI review costs?
GitHub estimates an average Copilot review at $0.25 for Lite, $1.50 for Balanced and $10 for Max. GitHub Actions minutes are additional.
Who is responsible for code an agent helps write?
GitHub says developers remain accountable for code that ships. Its cloud agent cannot approve or merge its own pull request, and the requester cannot approve that change.
Source: ThorstenMeyerAI.com
