🔍 Read the full analysis: AI Agents Start 4 In 10 GitHub Pull Requests, Raising Review Questions 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 2026 Octoverse report says more than 40% of pull requests in September started with a detectable coding agent, up from 17% in March. The figure is a floor, not a measure of code quality; GitHub also announced review tools and rules that keep final accountability with people.
More than 40% of pull requests on GitHub in September 2026 began with a detectable coding agent, up from 17% in March, according to the platform’s Octoverse report released October 28. GitHub says the September figure represents 110 million agent-started pull requests, a shift that makes reviewing and approving code a growing part of the software development workload.
GitHub defines an agent-started pull request as one whose first draft came from a coding agent the company can detect. The count includes public and private repositories unless a metric is specifically excluded. GitHub says undetected agent use means the reported share is a floor rather than a complete count. A pull request can still count as agent-started even if a person later revises much of the code.
The report says 90% of agent-started pull requests were merged within two weeks in September, compared with 83% in February. The corresponding rate for pull requests without a detected agent remained about 86%. Those figures describe whether changes were merged, not whether they were correct, maintainable or free of defects.
GitHub’s response includes Copilot code review Max, which uses several AI reviewers from different model providers and combines their feedback, and Copilot Fusion, which routes work to models for individual steps. The company also made Copilot review approvals eligible to count toward required approvals. GitHub estimates an average review costs $0.25 for Lite, $1.50 for Balanced and $10 for 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 a Bottleneck
The numbers point to a change in where development time may be spent: if agents produce more initial drafts, people still need to inspect changes, test them and decide whether to merge. A higher merge rate alone does not establish that the code is sound. Teams need measures such as rework, reversions and production incidents alongside counts of pull requests and approvals.
Review has a direct cost as well as a time cost. GitHub’s per-review estimates give organizations a way to model spending, but the listed prices do not include the additional cost of Actions minutes or the staff time needed to assess output. Companies will also need to decide which projects or change types can use AI review toward required approvals, and how much human scrutiny each warrants.
For developers, agents may take on small, routine tasks that once gave newer colleagues practice. The report does not measure that effect on training or hiring, but it makes the question relevant: teams may need deliberate ways to teach writing precise tasks and critically reading diffs when first drafts are increasingly automated.
Products Worth Considering
Bose Lifestyle Ultra Wireless Home Speaker
DJI Mic 3 Wireless Microphone System (2 Transmitters, 1 Receiver, Charging Case)
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
As an affiliate, we earn on qualifying purchases.
How GitHub Defines Agent-Started
The Octoverse report is GitHub’s annual review of activity on its platform; the 2026 edition covers October 1, 2025, through September 30, 2026. Its terminology varies across the report: the definition says “agent-started” and “first draft,” while some chart captions use “agent-generated” or “agent-authored.” GitHub also uses “agent-opened” in an industry section. The most precise wording for the headline statistic is that a detectable agent started the first draft.
GitHub says it recalculated 2024 and 2025 figures using updated definitions and a revised reporting window, so the historical numbers are not directly comparable with earlier Octoverse editions. The report also describes broader platform growth: it records 1.3 billion merged pull requests during the year, up from 538 million in 2025. That total provides scale for the platform but does not establish that agents caused the increase.
GitHub reports that 55% of Copilot monthly active users used agent features, compared with 35% a year earlier, and says use of its cloud agent grew 230% year over year. These are company-reported platform metrics. They show adoption, not the quality or productivity impact of agent-written code.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in its Octoverse definition
What the Adoption Figure Cannot Show
GitHub does not publish a public-versus-private repository breakdown for the 40%-plus figure, nor a full account of which agents its detection catches. The actual share of pull requests that began with any coding agent may be higher, but the report does not quantify how much higher. Its shifting labels also make it important to preserve the stated definition when quoting the statistic.
The report does not establish whether agent-started pull requests cause more defects, require more reviewer time, or improve delivery speed. The higher two-week merge rate does not answer those questions: merged is not the same as correct. GitHub’s price estimates likewise do not provide a complete cost comparison across review modes or an independent assessment of review accuracy.
It remains unclear how organizations will set approval policies for AI-generated reviews, or how often teams will rely on them for consequential changes. The report does not provide a breakdown of incidents, reversions or rework tied to agent-started code. Those outcomes will be needed to judge whether faster code production yields a net benefit.
Measure Quality Alongside Volume
GitHub’s new review and routing products give users options for handling agent-written changes, but the report does not set a timetable for adoption or provide results from a broad evaluation of their accuracy. Companies using the tools will need to decide where AI feedback can count toward approvals, how to allocate review credits and reviewer time, and when a person must perform an independent review.
The next useful evidence will be operational: whether agent-started changes need more revisions, are reverted more often, or contribute to production problems. GitHub has not published those comparisons in the report. Until it does, organizations should treat pull-request volume and merge rates as measures of activity and throughput, not as proof of code quality or successful deployment.
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 can revise the draft later and it still counts. The company says its detection misses some agent use.
Did more than 40% of GitHub code come from AI?
No. The figure concerns the share of pull requests that began with a detectable agent, not the share of all code or commits written by AI. GitHub does not provide a full measure of agent-written code in this statistic.
Does the higher merge rate show that agent code is better?
No. GitHub says 90% of agent-started pull requests were merged within two weeks in September, but merging does not establish correctness or quality. The report does not compare defect, rework or incident rates for the two groups.
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. These estimates are in addition to GitHub Actions minutes.
Who is responsible for code an agent produces?
GitHub says developers remain accountable for what ships. Its cloud agent cannot approve or merge its own pull request, and GitHub says the person who requested a change cannot approve it either.
Source: ThorstenMeyerAI.com
