🔍 Read the full analysis: The Changing Balance Of GitHub Pull Requests: AI Starts 4 In 10 on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
GitHub’s Octoverse 2026 report says more than 40% of pull requests on the platform in September began with a coding agent’s first draft, up from 17% in March. The company says the count is a floor because it cannot detect every agent, and a merged pull request is not proof that code is correct. GitHub introduced tools for AI-assisted review and routing, while keeping responsibility for shipped code with developers.
Coding agents started more than 40% of pull requests on GitHub in September, up from 17% in March, according to the platform’s Octoverse 2026 report, released October 28. GitHub counted 110 million agent-started pull requests in September, a rapid rise that shifts attention from generating code to reviewing and approving it.
GitHub defines an agent-started pull request as one whose first draft came from a coding agent the company can detect. It says its observed count is a floor because some agent use may go undetected. The reported share includes public and private repositories unless otherwise specified; GitHub did not provide a public-versus-private breakdown for this figure. The measure also does not mean an agent wrote every line: a pull request still counts if a person later substantially edits the draft.
The report 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 whether a change was merged, not whether it was accurate, safe, or free of later rework. Agent-started pull requests were also more likely to touch multiple files: 87% changed more than one file, compared with 79% of pull requests without a detected agent.
GitHub announced review and routing products alongside the report. Copilot code review Max combines reviews from several AI reviewers using different model providers. The company also said Copilot review approvals can count toward required approvals, and introduced Copilot Fusion, a system that routes tasks to models. GitHub estimates an average review costs $0.25 in Lite, $1.50 in Balanced, or $10 in 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 Bottleneck
The figures show that more pull requests are being started by coding agents, while teams still need to review, test and decide whether to accept proposed changes. A larger flow of proposed changes can increase demands on reviewers, especially when agent-started pull requests affect several files. GitHub’s report and product announcements address review capacity and approval rules alongside code generation.
The rising merge rate needs careful interpretation. A merge records an accepted change, but does not establish code quality or business value. Teams assessing agent use can also track measures such as rework, reversions and production incidents. GitHub’s pricing estimates show that review costs vary by mode, with Actions usage charged separately.
GitHub says developers remain accountable for what ships. Its cloud agent cannot approve or merge its own pull request, and the person who requested the change cannot approve it either. Organizations setting policies for agent-assisted work will need to determine who may approve changes and what review is required.
Products Worth Considering
Logitech MX Mechanical Wireless Illuminated Keyboard Tactile (Graphite)
LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
How GitHub Defines Agent Work
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 the agent share rose from about one in six pull requests in March to more than four in ten in September. It describes September’s volume as about six times March’s, while noting that the report’s detection-based methodology does not capture all agent use.
GitHub also reported a broader increase in activity: 1.3 billion pull requests were merged over the reporting period, compared with 538 million in 2025 and 402 million in 2024. It recalculated 2024 and 2025 figures using updated definitions and the reporting window, so the older Octoverse totals are not directly comparable with previously published versions. The report says use of Copilot’s cloud agent grew 230% year over year, and 55% of monthly active Copilot users now use agent features, compared with 35% a year earlier.
The report’s wording varies among “agent-started,” “agent-generated,” “agent-authored” and “agent-opened.” Its definition centers on a first draft from a detectable agent, making “agent-started” the clearest description of what the headline figure measures. It does not mean every such change was written solely by an agent or that every coding agent on the platform was counted.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in its Octoverse methodology
Limits of the September Count
The full share of pull requests involving agents is unknown. GitHub says undetected agent use is omitted, but has not quantified how much is missing. It also has not published a split of the 40%-plus figure between public and private repositories, or enough detail to determine how the rate varies by team, project or type of work.
The reported merge rate does not settle whether agent-started code is reliable. GitHub’s figures do not, by themselves, show defect rates, security outcomes, review hours, later rework or the share of merged changes that were subsequently reverted. The report’s differences in terminology also mean that readers should not treat “agent-authored” as proof that an agent wrote an entire pull request.
It is also unclear how often teams will use Copilot reviews as required approvals or how those reviews compare with human review for catching consequential problems. GitHub supplied estimated prices for its review modes, but actual costs will depend on usage and applicable Actions minutes.
Teams Weigh Review and Cost
The report provides a September snapshot. Future comparisons could show whether agent-started pull requests continue to grow and whether review and quality outcomes change. GitHub’s next Octoverse reporting could provide a later platform-wide measure, while companies can monitor their own repositories for review time, rework, reversions and production incidents.
Organizations adopting the new tools will also have to set policies for when AI review can count toward an approval requirement, who is permitted to approve a change, and what additional checks are needed before code ships. The pricing estimates provide per-review tier costs, but final spending will depend on actual review volume and Actions use. GitHub has not provided a date for a subsequent report or further outcome data tied to these new controls.
Key Questions
What does GitHub mean by an agent-started pull request?
GitHub counts a pull request when its first draft came from a coding agent it can detect. A person may edit the draft later and it still counts. GitHub says the measure misses some agent use.
Does the 40% figure mean agents wrote 40% of all code on GitHub?
No. It refers to the share of pull requests in September that GitHub detected as starting with an agent’s first draft. It is not a measure of the share of all code written by agents.
Does merging an agent-started pull request mean its code is correct?
No. The reported merge rate shows that pull requests were merged within a given period. It does not establish correctness, safety, or the absence of later rework or incidents.
How much does GitHub say an AI code 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 produced by an agent?
GitHub says developers remain accountable for what ships. Its stated controls prevent the cloud agent from approving or merging its own pull request, and prevent the person who requested a change from approving it.
Source: ThorstenMeyerAI.com
