🔍 Read the full analysis: AI Agents Start 4 In 10 GitHub PRs: How Teams Can Adapt Their Reviews on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
GitHub’s Octoverse report says more than 40% of pull requests on its platform started with a detectable coding agent in September 2026, up from 17% in March. The company introduced review and model-routing tools as agent-written changes grow, but its merge data does not establish code quality, and the report leaves important measurement details open.
More than four in 10 GitHub pull requests started with a coding agent’s first draft in September, up from 17% in March, according to GitHub’s Octoverse report released October 28. The company also introduced AI review and model-routing tools, reflecting a shift in the software workflow: as agents draft more proposed changes, teams must decide how to inspect, approve and pay for them.
GitHub says the September total was 110 million agent-started pull requests, about six times the March volume. It defines an agent-started pull request as one whose first draft came from a coding agent the platform can detect. GitHub warns that its count is a floor because some agent use may go undetected. The figure can still include a pull request that a person substantially rewrote after the agent’s initial draft.
In September, 90% of detected agent pull requests were merged within two weeks, compared with 83% in February. Pull requests without a detected agent remained at about 86%. These figures describe whether changes were merged, not whether they were correct, safe or free of later rework. GitHub also reports that 87% of agent pull requests changed more than one file, versus 79% of those without a detected agent.
GitHub’s new offerings focus on review and workflow control. Copilot code review Max uses several AI reviewers from different model providers and combines their findings; Copilot approvals can count toward required approvals; and Copilot Fusion routes models to different steps. GitHub estimates an average review costs $0.25 for Lite, $1.50 for Balanced and $10 for Max, plus GitHub Actions minutes. Those are company estimates, not an independent assessment of total review cost.
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 growth changes where teams may spend engineering time. If agents make it faster to produce a first draft, the work of reading diffs, checking behavior and deciding whether to merge can become the limiting step. Agent changes that touch several files may also demand more reviewer attention, though file count alone does not show that a change is harder or riskier.
Teams adopting these tools face practical choices about reviewer time, AI review budgets and approval policy. They can set rules for which changes need human review, whether an AI review may count toward formal approvals, and how to track costs. The useful measures are not just the number of pull requests opened or merged: teams can also monitor rework, reversions and production incidents to see whether speed comes with acceptable outcomes.
There may be workforce effects as well. Small, bounded coding tasks have often provided early-career developers with practice. If agents take on more of that work, managers may need to create other opportunities for junior staff to learn how to interpret requirements, test changes and review code. The source material raises this as a concern, but does not quantify how widely it is happening.
Popular right now
Breville BES870XL Barista Express Espresso Machine, Brushed Stainless Steel
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
As an affiliate, we earn on qualifying purchases.
How GitHub Defines Agent Work
The Octoverse 2026 report covers October 1, 2025, through September 30, 2026. GitHub’s methodology says platform metrics include public and private repositories unless otherwise specified; the report does not provide a public-versus-private breakdown for the agent share. GitHub also recalculated its 2024 and 2025 figures using updated definitions and reporting windows, so comparisons with older Octoverse reports may not be direct.
The report describes broader growth on the platform: GitHub recorded 1.3 billion merged pull requests in the year ending in September, compared with 538 million in 2025 and 402 million in 2024. Those platform-wide totals provide scale, but they do not isolate how much of the increase came from agents. GitHub says agent use rose as tools moved from single-session assistance toward persistent cloud agents that can work on assigned tasks over time.
“A pull request whose first draft came from a coding agent that we can detect.”
— GitHub, in its Octoverse report
What the Metrics Cannot Show
The report does not say what share of agent-started pull requests came from public versus private repositories, nor does it provide an independent audit of whether agents were detected consistently. Since undetected use is excluded, the reported percentage should not be treated as a complete count of agent involvement.
GitHub’s wording also varies across the report, using terms such as “agent-started,” “agent-generated,” “agent-authored” and “agent-opened.” The most specific definition is that the first draft came from a detectable agent; the figures do not mean an agent wrote every line that was eventually merged. And a merged pull request is not evidence by itself that the code was correct or caused no later problems. The report provides no comprehensive quality comparison or outcome data establishing that agent changes are better or worse.
How Teams Can Set Review Rules
GitHub’s report and product launches put review policy and cost tracking on the agenda for teams using coding agents. Organizations can decide which changes require a human sign-off, when AI findings are useful, and whether an AI approval can satisfy any required review step. They can also compare review-tool spending with subsequent rework, reversions and incidents rather than treating faster merges as proof of success.
The next evidence to watch is whether GitHub publishes further breakdowns of agent use and whether teams report measured quality outcomes alongside adoption. Until those details are available, the September figures establish a rapid rise in detectable agent-started work, but leave its full reach and effect on software quality unresolved.
Key Questions
What does GitHub mean by an agent-started pull request?
GitHub defines it as a pull request whose first draft came from a coding agent it can detect. A person may have revised the draft before submission or merge.
How many GitHub pull requests started with agents in September?
GitHub reports 110 million in September 2026, representing more than 40% of pull requests on the platform that month. The company says the count is a floor because some agent use may go undetected.
Does the merge rate show that agent-written code is reliable?
No. The reported 90% two-week merge rate measures whether pull requests were merged. It does not establish correctness, security or the absence of later rework.
How much does GitHub say AI code review costs?
GitHub estimates an average review at $0.25 for Lite, $1.50 for Balanced and $10 for Max, plus GitHub Actions minutes. These are GitHub’s estimates, not a full independent accounting of costs.
Who is responsible for code an agent helps produce?
GitHub says developers remain accountable for what ships. It says its cloud agent cannot approve or merge its own pull requests, and the person who requested the change cannot approve it either.
Source: ThorstenMeyerAI.com
