
We built a Coding AI Leaderboard that tracks bot activity on GitHub using the public GH Archive dataset. Let's take a look.
AI coding assistants — CodeRabbit, Copilot, Gemini Code Assist, Cursor, ChatGPT Codex, and others — now account for over 60% of all bot PR reviews on GitHub, up from roughly 20% at the start of 2025.

That's one in five to two in three in about a year. The biggest jump came after Sonnet 3.7 / GPT-4.5 launched in February 2025 — penetration roughly doubled from ~20% to ~40% within weeks.
Here's the most recent 30-day window (Feb 9 – Mar 10, 2026).

CodeRabbit leads with 179,965 reviews in 30 days — nearly double Copilot's 91,596. Among the AI coding tools further down: ChatGPT Codex Connector (#4 at 47,694), Gemini Code Assist (#6 at 43,520), Cursor (#9 at 27,250), Devin (#10 at 19,953), Greptile (#11 at 17,693), and Claude (#12 at 12,002). The remaining spots are non-AI bots like GitHub Actions and Amazon Inspector.
CodeRabbit overtook Copilot for the #1 spot by mid-2025 and has held it since. The mid-tier keeps shuffling — Gemini Code Assist, ChatGPT Codex Connector, and Cursor trade places quarter to quarter.


Dependency bots still lead PR creation — Dependabot (403K), Pull (272K), and Renovate (120K) hold the top three spots. But Copilot is #4 with 88,943 PRs opened — an AI coding assistant ahead of GitHub Actions.
Devin (#13 at 2,847) and Google Labs Jules (#16 at 2,550) are here too — AI agents that write and submit code, not just review it. Coding AI went from near 0% to about 9-10% of all bot PRs over the past year.
Caveat: these numbers only count PRs where a bot account is the author. Many PRs are submitted under a human's GitHub account but actually written by AI — think 🤖 Generated with Claude Code or Cursor-authored commits pushed by the developer. The real number is higher.


The pushes leaderboard is where AI app builders show up. Lovable Dev ranks #5 with 281,309 pushes — you describe what you want, it builds the app and pushes to a repo. Base44 Builder (#20 at 25,534) does the same thing.
Coding AI pushes climbed from near zero to about 2-3% of all bot pushes, with recent days above 5%.

The one area where AI coding tools haven't made major inroads is issue creation. AI accounts for less than 1.5% of bot-created issues, and the growth has been uneven. This makes sense — filing useful issues requires understanding project context, user intent, and bug triage in a way that's harder to automate than code review or generation.

Looking at the last 30 days table, Claude appears at #18 with 2,112 issues, the only major AI coding assistant meaningfully present in this category.

Linear (#10 at 3,257) has been climbing fast in the ranking chart. Linear isn't an AI coding tool itself — it's a project management system that syncs issues bidirectionally with GitHub. But its rapid rise hints at a growing agentic workflow: teams create issues in Linear, which sync to GitHub, where AI coding agents pick them up and open PRs.

The penetration charts mark major model releases as red dashed lines. Correlation isn't causation — product launches, pricing changes, and seasonal effects all play a role — but a few jumps stand out:
PRs Opened is different: near zero through the first half of 2025, then a gradual ramp to ~10% by late 2025. The day-to-day noise is high, and the growth looks more like steady adoption than anything tied to specific releases.
AI bots already handle most bot code review on GitHub. They're starting to write the code too. And remember — this only counts bot accounts. Every developer pushing AI-written code under their own name doesn't show up here. The numbers on this page are a floor, not a ceiling.
Explore the full data on the Coding AI Leaderboard.