1 in 7 PRs now involve AI agents— 14X since early 2024
Based on AI review activity observed across 40.3M pull requests
AI Agent Participation
AI code reviews went from rare to mainstream in three years#
AI participation grew from under 1% in 2022 to over 14% today, with steady adoption across thousands of organizations. This reflects a structural shift in how code is reviewed.
The Long Arc of Adoption
Different metrics surfaced different leaders#
CodeRabbit led by pull request volume, while Copilot dominated organizational adoption. Leadership shifted depending on whether impact was measured by activity or reach.
Agent Leaderboard
| Rank | Agent | Pull Requests | Total Events |
|---|---|---|---|
| 1 | CodeRabbit | 632,256 | 2,706,240 |
| 2 | Copilot | 561,382 | 1,934,836 |
| 3 | Gemini | 174,766 | 542,368 |
| 4 | Sourcery AI | 131,900 | 314,422 |
| 5 | ChatGPT | 87,617 | 159,614 |
| 6 | Greptile | 54,564 | 220,387 |
| 7 | Amazon Q | 49,563 | 430,454 |
| 8 | Cursor | 48,770 | 129,747 |
| 9 | Claude | 24,916 | 58,230 |
| 10 | Ellipsis | 24,516 | 67,441 |
Ranked by distinct PRs each agent participated in during 2025.
Platform giants dominated market adoption#
Copilot’s native GitHub integration led org adoption. ChatGPT overtook Gemini in adoption despite Gemini’s PR momentum. CodeRabbit held a strong #2 position against the giants.
Cumulative Org Adoption by Agent (2025)
Note: Org adoption is based on agent activity in public repositories only.
A long tail of AI agents emerged, and some churned#
Beyond platform leaders and high-traction tools, dozens of AI agents showed measurable PR activity. At the same time, some early entrants, including Korbit, exited the market, reflecting rapid experimentation and turnover.
Long-Tail AI Agents by Adoption
Get the Full Report
Enter your email to receive the PDF report with methodology, full tables, and the complete agent ranking.
Early access: Get the full PDF the moment it's released. Unsubscribe anytime.
Methodology
What this report measures
A pull request counts as AI-participated if at least one AI agent submits a review or comment on that PR.
Data source
- •Source: GitHub Archive + PullFlow Anonymized
- •Scope: Public GitHub events (reproducible from the same source)
- •Events: Reviews, review comments, and issue comments on pull requests
- •Time range: 2022–2025 snapshots; monthly in 2025
Repository selection
To focus on meaningful code review activity, we filter to repositories that meet:
Agent identification
An AI code review agent is a bot account that uses an LLM to analyze code or diffs and creates review or issue comments on pull requests.
Citation and Revisions
Revision History
Corrected data errors found in the initial release:
- Corrected Gemini growth calculation (now 12X based on distinct orgs).
- Fixed cumulative adoption chart scale to reflect unique organizations.
- Updated AI authorship volume to include missed activity (335k+ PRs).
- Expanded long-tail agents chart to include mid-sized agents.
Connect GitHub, Slack, IDEs, and AI agents
for unified code review.
"PullFlow takes GitHub PR collaboration to the next level — seamless, efficient, and built for real developer workflows."
Tom Preston-Werner · Co-Founder, GitHub