🔍 Read the full analysis: How AI Agents Are Reshaping Pull Request Review On GitHub 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 its platform in September began with an AI coding agent’s first draft, up from 17% in March. GitHub also introduced review and model-routing tools, while saying developers remain accountable for code that ships. The figures measure detected agent activity, and merged pull requests are not necessarily correct or high-quality.
More than 40% of pull requests on GitHub in September started with a coding agent’s first draft, up from 17% in March, according to the company’s Octoverse 2026 report. Released October 28 at GitHub Universe, the report comes alongside new tools for AI-assisted review, as the company says developers remain responsible for code that is approved and shipped.
GitHub says the September share represents 110 million agent-started pull requests in one month, roughly six times the March volume. It defines an agent-started pull request as one whose first draft came from a coding agent the company can detect. GitHub describes the count as a floor because some agent use may not be detected. The metric includes public and private repositories, but the company did not provide a separate breakdown for them.
The company reported that 90% of detected agent pull requests were merged within two weeks in September, up from 83% in February. The comparable rate for pull requests without a detected agent was about 86%. Those figures describe whether changes were merged, not whether they were correct, safe or free of later rework. GitHub also said 87% of agent-started pull requests changed more than one file, compared with 79% of other pull requests.
GitHub’s launches focus on handling review volume. Copilot code review Max uses reviewers based on different model providers and combines their feedback; the company also made Copilot approvals eligible to count toward required approvals. Its new Copilot Fusion routes tasks to models for different steps. GitHub estimates an average review costs $0.25 on Lite, $1.50 on Balanced and $10 on Max, plus GitHub Actions minutes.
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 AI assistance is changing where time is spent in software development: agents can produce initial changes quickly, but a person still has to judge whether those changes should be accepted. As pull request volume rises, reviewer attention and approval workflows can become bottlenecks, even when code drafting takes less time.
For engineering teams, the new review options also introduce a budgeting question. GitHub’s quoted per-review estimates make AI review a metered cost, in addition to the time developers spend checking results and any Actions usage. Companies will need to decide where AI feedback helps, when a human must give the required approval, and how to track outcomes such as rework, reversions and incidents rather than treating merged pull requests as proof of quality.
The shift may affect developers’ work as well. Writing clear task instructions and carefully reading changes become more important when an agent produces the first draft. Entry-level tasks have often given junior developers practice with code and review; the report does not measure how hiring or training is changing, but greater automation of routine work makes those questions relevant to teams.
Products Worth Considering
EcoFlow Smart Home Panel 2 Whole-Home Backup System Hub
Logitech MX Brio Ultra HD 4K Webcam, Graphite
LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How GitHub Defines Agent Work
Octoverse is GitHub’s annual report on activity across its platform. The 2026 edition covers October 1, 2025, through September 30, 2026. GitHub’s agent count is based on signals it can detect, rather than a complete audit of every developer’s tool use. A pull request still counts as agent-started if a person later rewrites part of the draft.
The report uses several nearby labels, including “agent-started,” “agent-generated” and “agent-authored.” Its stated definition centers on the first draft, so agent-started is the most precise description of what the headline percentage measures. GitHub also recalculated its 2024 and 2025 figures using updated definitions and a changed measurement window, meaning earlier Octoverse totals are not directly comparable.
The growth occurred amid a broader rise in platform activity. GitHub reported 1.3 billion merged pull requests during the report year, compared with 538 million in 2025 and 402 million in 2024. It also said 55% of monthly active Copilot users used agent features, up from 35% a year earlier. These are company-reported platform metrics, not an independent assessment of code quality or productivity.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in its Octoverse 2026 report
What the Metrics Cannot Show
GitHub has not published a split of the agent-started figure between public and private repositories, nor a complete account of agent use that its systems fail to detect. Its changing labels also make precise interpretation important: the measure concerns an agent’s role in the first draft, not the share of final code written by an agent.
The report’s merge rate does not establish that agent-written changes are as reliable as human-written changes. GitHub has not supplied, alongside these figures, a comparison of defect rates, security issues, reversions or long-term maintenance costs. It is also unclear how often teams accept AI review feedback or how Copilot approvals affect review quality in practice. The reported prices are estimates, and actual costs may vary with usage and Actions minutes.
Teams Weigh Review Costs and Controls
GitHub’s new review and routing features give teams options for handling agent-written changes, but the company’s report does not establish how widely they will be adopted or what results they will produce. The next useful evidence will be measured outcomes over time: whether teams find defects sooner, reduce rework, or see changes in incidents and review effort as agent use grows.
For now, organizations face decisions about which repositories can use AI review, when a human approval is mandatory, and how to budget review credits and Actions usage. Developers will continue to review proposed changes under GitHub’s stated accountability rules. Whether the tools make that work faster without weakening scrutiny remains to be seen.
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 that the company can detect. A person may later edit or rewrite the draft and it still counts.
Did AI agents start 40% of all code changes on GitHub?
No. GitHub’s report says more than 40% of pull requests in September 2026 began with an agent’s first draft. It does not say that agents wrote 40% of all code, and undetected agent use is not included.
Does a high merge rate mean agent-written code is reliable?
No. GitHub reported that 90% of detected agent pull requests were merged within two weeks, but merging is not a measure of correctness. The report does not provide a corresponding defect or incident comparison.
How much does Copilot code review cost?
GitHub estimates an average review at $0.25 for Lite, $1.50 for Balanced and $10 for Max, plus GitHub Actions minutes. Actual costs may depend on usage.
Who is responsible for code an AI agent proposes?
GitHub says developers remain accountable for what ships. The company says its cloud agent cannot approve or merge its own pull request, and the person who requested the change cannot approve it either.
Source: ThorstenMeyerAI.com
