4 in 10 GitHub pull requests now start with an AI agent

More than 40% of pull requests on GitHub in September started with an AI agent’s first draft.

The numbers

1. September 202640%+

   of pull requests on GitHub started with an agent’s first draft.

   In March it was 17%. That is six months.

   Source: GitHub Octoverse 2026, 28 Oct 2026

2. One month110M

   agent-started pull requests in September alone.

   About six times March’s volume.

   Source: GitHub Octoverse 2026, 28 Oct 2026

3. Merged PRs with AI involvement

   Among larger languages, TypeScript leads

   • TypeScript44.1%
   • Python41.1%
   • Rust40.4%
   • Java21.2%

   Only AI that GitHub could detect is counted.

   Source: GitHub Octoverse 2026, 28 Oct 2026

4. Repos created in 20268x

   more commits in repos that have an agent config file.

   Correlation, not cause: busy repos may add the file first. Still, the file is cheap.

   Source: GitHub Octoverse 2026, 28 Oct 2026

5. 12 months to 30 Sep 202617.9B

   pushes to GitHub. More than 2024 and 2025 combined.

   Source: GitHub Octoverse 2026, 28 Oct 2026

Three moves for your team

1. Write your rules where the agent reads them. Conventions, test commands, what never to touch. One file in the repo.
2. Watch review, not just output. More drafts means more to read. Track how long PRs wait for a human.
3. Treat 40% as a floor. GitHub counts only the agents it can detect. The real share is higher.

Sources

1. GitHub Octoverse 2026, 28 Oct 2026

What share of your PRs start with an agent? Tell me on LinkedIn or Instagram.