🔍 Read the full analysis: What A 4-In-10 Share Of AI-Started GitHub PRs Means For Review on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- 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 on its platform in September 2026 began with a detectable coding-agent draft, up from 17% in March. The figures measure how a pull request started, not whether its code was correct; GitHub says 90% of detected agent PRs were merged within two weeks. The increase puts review capacity, review-tool costs and responsibility for approving shipped code in focus.
More than four in 10 GitHub pull requests in September began with a first draft from a detectable coding agent, up from 17% in March, according to GitHub’s Octoverse 2026 report, released October 28. The shift matters because generating a proposed change is becoming faster while people still need to inspect, evaluate and approve the code before it is merged.
GitHub reports 110 million agent-started pull requests in September and says the number was about six times March’s volume. Its definition is specific: a pull request counts when its first draft came from a coding agent GitHub can detect. The company calls the measured share a floor, because some agent use may not register in its data. The report includes public and private repositories unless otherwise stated, but does not provide a public-versus-private breakdown for this figure.
GitHub also says 90% of detected 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 figures describe merge rates, not code quality: a merge does not establish that a change was correct, safe or free from later rework.
The company’s product announcements at its Universe conference focused in part on review. Copilot code review Max brings together several AI reviewers using different model providers and produces a combined review. GitHub also said Copilot approvals can count toward required approvals, and introduced Copilot Fusion, a system that routes work to models for different steps. GitHub estimates an average Copilot 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 Capacity Becomes the Constraint
The figures suggest that a growing portion of proposed code may arrive faster than teams can evaluate it. Agent-started changes are not necessarily larger in every sense, but GitHub says 87% change more than one file, compared with 79% of pull requests without a detected agent. A reviewer may need to follow how several files interact, check whether the request was fulfilled and decide whether tests and safeguards are adequate.
That makes review a question of staff time and operating cost, not just developer convenience. AI review tools add a per-review charge and may also consume Actions minutes. Teams will need to decide which changes warrant more intensive review, who can approve them, and how to monitor outcomes such as reverts, defects and incidents rather than treating a high merge count as evidence of success.
The shift also affects individual developers. When agents produce first drafts, writing a clear task and critically reading a diff become more central parts of the work. The source report raises a further workforce concern: small, introductory coding tasks that have often helped junior developers learn may be among the tasks agents can take on first. It does not establish how hiring or training will change, but the changing task mix is relevant to teams planning how new developers gain experience.
Products Worth Considering
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
Apple iPad Air (2025) 13‑inch Wi‑Fi 256 GB – Space Gray (M3 chip)
As an affiliate, we earn on qualifying purchases.
How GitHub Defines Agent-Started
The numbers come from Octoverse 2026, GitHub’s annual report on activity across its platform. This edition covers October 1, 2025, through September 30, 2026. GitHub defines an agent-started pull request as one whose first draft came from a coding agent it can detect. A request still counts under that definition even if a person later rewrites part of the draft.
The report uses several related labels, including “agent-started,” “agent-generated,” “agent-authored” and “agent-opened.” For the headline share, agent-started is the most precise description of GitHub’s stated measure. The company also recalculated its 2024 and 2025 figures using new definitions and reporting windows, so earlier Octoverse totals are not directly comparable without accounting for those changes.
GitHub’s report describes a broader rise in activity: it records 1.3 billion merged pull requests over the year, up from 538 million in 2025 and 402 million in 2024. That platform-wide growth is background, not proof that agents alone caused the increase. The report says agent use rose as tools developed from single-session assistants toward persistent cloud agents that can work across days.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in the Octoverse 2026 report
What the Merge Figures Cannot Show
GitHub has not disclosed how the 110 million September pull requests divide between public and private repositories, or how many came from particular agents. Its detection method also leaves some agent use uncounted, while the report’s varying labels can make comparisons harder to interpret. The figures do not show how much of each draft a human changed before submitting or merging it.
Most important, merge rates are not quality measures. The report does not establish whether agent-started changes caused more defects, rework, security problems or later reversions than other changes. Nor does it provide evidence that AI review tools catch those problems at a particular rate. GitHub says its cloud agent cannot approve or merge its own pull requests, and that the person requesting a change cannot approve it; the exact approval rules and their application across different workflows still warrant attention.
Track Quality Alongside Adoption
GitHub’s announced review products put pricing and approval policies into practical use, but the report does not set out a universal standard for when an AI review should count toward a required approval. Organizations adopting the tools will need to set their own rules for which changes can use automated review, when a human reviewer must remain in the approval path, and how to budget for review tiers and Actions usage.
The next useful evidence will be outcome data over time: defect rates, reverts, review time and incidents, broken out by the kind of work and review process. Until comparable results are available, the September share shows how frequently detectable agents started pull requests—not whether those changes were better, cheaper or safer to ship. GitHub’s future reporting and customers’ own measurements will help clarify that distinction.
Key Questions
What does “agent-started pull request” mean?
GitHub counts a pull request when its first draft came from a coding agent the company can detect. A person may edit the draft later and it still meets that definition.
Does the 40% figure mean 40% of GitHub code was written by AI?
No. It refers to the share of pull requests that started with a detectable agent draft. It does not measure the share of all code written by AI, or how much of each submitted change was later rewritten by a person.
Were agent-started pull requests more likely to be merged?
GitHub says 90% were merged within two weeks in September, compared with 83% in February. Pull requests without a detected agent were at about 86%. Merge rates do not establish correctness or quality.
How much does GitHub estimate AI review costs?
GitHub estimates an average review at $0.25 for Lite, $1.50 for Balanced and $10 for Max. GitHub Actions minutes are an additional cost.
Who is responsible for code an agent drafts?
GitHub says developers remain accountable for what ships. The company says its cloud agent cannot approve or merge its own pull requests and that the person who requested a change cannot approve it either.
Source: ThorstenMeyerAI.com
