🔍 Read the full analysis: 4 In 10 GitHub PRs Start With AI Agents: A Look At The Review Bottleneck 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 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’s Octoverse report also shows high merge rates for agent-started PRs, but a merge does not establish that code is correct. GitHub announced new AI review and routing tools as human review capacity and approval rules become central concerns.
More than 40% of GitHub pull requests in September 2026 began with a first draft from a coding agent GitHub could detect, the company said in its Octoverse report released October 28. The share was 17% in March, according to GitHub, a rise that puts pressure on the people and systems responsible for reviewing code before it is merged.
GitHub reported 110 million agent-started pull requests in September. It defines an agent-started PR as one whose first draft came from a coding agent the company can detect. GitHub says the figure is a floor because its methods may miss some agent use. It counts a PR as agent-started even if a person later changes part of the draft.
In September, 90% of detected agent-started PRs were merged within two weeks, compared with 83% in February. PRs without a detected agent remained at about 86%, GitHub reported. Those figures describe whether a change was merged, not whether it was error-free, well-designed or free of later rework.
GitHub’s product announcements at its Universe conference put attention on AI agent interfaces, review and approval. The company introduced Copilot code review Max, which uses several AI reviewers from different model providers and combines their findings; made Copilot approvals eligible to count toward required approvals; and released Copilot Fusion, a system that routes tasks to models. GitHub estimates an average review will cost $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 reported shift changes where time may be spent in software development. If agents produce more initial drafts, the work of reading diffs, checking behavior and deciding whether to approve can become a larger share of a team’s workload. GitHub’s figures suggest the increase is not limited to a small number of experimental PRs, though its detection method means the precise scale is uncertain.
For engineering organizations, the new review options also introduce a budget and governance decision. GitHub’s estimated prices vary widely by review tier, and the service uses additional Actions minutes. Teams will need to decide which changes merit an AI review, whether an AI-generated approval can satisfy a required review rule, and how to judge value through rework, reversions and production incidents rather than PR counts alone.
For developers, generating code may take less time while writing precise tasks and scrutinizing proposed changes become more important. Smaller, routine tasks have also been a way for less experienced developers to learn a codebase. The report does not establish how widely that training role is changing, but the increased use of agents makes it a practical issue for teams to monitor.
Products Worth Considering
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How GitHub Defines Agent-Started
The Octoverse report covers the year from October 1, 2025, through September 30, 2026. GitHub’s methodology says its metrics include public and private repositories unless a metric is marked otherwise; the agent-share figure has no stated public-versus-private breakdown. GitHub did not provide a split between open-source and company repositories for this measure.
The report uses several nearby terms, including “agent-started,” “agent-generated” and “agent-opened.” Its 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. GitHub also recalculated 2024 and 2025 figures using updated definitions and reporting windows, meaning older Octoverse figures are not directly comparable without that adjustment.
The agent data appeared alongside a broader rise in GitHub activity. The report says 1.3 billion pull requests were merged over the 12 months ending in September, up from 538 million in 2025. GitHub also reported that 87% of agent-started PRs changed more than one file, compared with 79% of PRs without a detected agent. That difference may mean more material for reviewers to assess, but the report does not measure review time or code quality.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub’s Octoverse report
What the Metrics Do Not Show
GitHub has not published the public-versus-private repository breakdown for the agent share, nor a separate figure for company code and open-source projects. Because the count depends on detectable signals, the actual share of PRs that began with agent assistance could differ from the reported percentage. The report’s varying labels also make it important to retain the specific definition when citing the figure.
The merge-rate data does not show whether changes introduced bugs, required substantial human revision, or were later reverted. GitHub has not supplied a comparison of review time, downstream defects or rework for agent-started and other PRs. Its listed review prices are estimates, and the source material does not provide a detailed accounting of actual costs across teams or workloads.
GitHub says its cloud agent cannot approve or merge its own pull requests, and that the person who requested a change cannot approve it. The source does not explain how organizations will configure Copilot approvals across different repository rules or how often teams will rely on AI review in place of a human reviewer.
Teams Set Review Rules and Budgets
The immediate next step for teams adopting these tools is to decide how they fit into existing approval policies, review workloads and software budgets. GitHub’s new review options provide more ways to automate feedback, but organizations will need to test whether those reviews catch useful issues and whether their costs fit the work being checked.
GitHub’s report offers a platform-wide snapshot, not a forecast of how adoption will develop. Future comparisons will be clearer if the company reports consistent definitions alongside quality and outcome measures, including rework and incidents. Until then, the September share and merge rates establish the scale GitHub is reporting, while leaving the effect on software quality and human review time unsettled.
Key Questions
What does it mean that 4 in 10 GitHub PRs started with AI agents?
GitHub says more than 40% of pull requests in September 2026 had a first draft from a coding agent it could detect. The figure does not mean that agents wrote every line in those PRs; people may have revised the drafts.
How many agent-started pull requests did GitHub report?
GitHub reported 110 million in September 2026. It says the count is a floor because its systems may not detect every instance of agent use.
Does a high merge rate show that agent-written code is reliable?
No. GitHub said 90% of detected agent-started PRs were merged within two weeks, but merging is not a measure of correctness. The report does not provide comparative defect, rework or reversion rates.
What will GitHub’s AI code reviews cost?
GitHub estimates average reviews at $0.25 for Lite, $1.50 for Balanced and $10 for Max, plus GitHub Actions minutes. The figures are estimates, not a report of actual spending by customers.
Who is accountable when an AI agent’s code ships?
GitHub says developers remain accountable. The company also says its cloud agent cannot approve or merge its own PR, and the person who requested a change cannot approve it.
Source: ThorstenMeyerAI.com
