I analyzed 40.3 million public pull requests from 2022-2025.
The data is brutal: GitHub Copilot now dominates organizational adoption despite CodeRabbit processing more total PRs.
This isn't a hot take. It's what 40M PRs told us.
The Uncomfortable Truth
CodeRabbit was built from the ground up for AI code review. It's genuinely good at what it does.
👑 CodeRabbit: #1 in PR volume for 2025.
But that crown is slipping. Copilot overtook them in monthly PRs in November, and already leads in org adoption.
Copilot wasn't built for code review. It started as autocomplete. Yet here we are.
Why Platform Usually Wins
Here's what the data actually shows:
Copilot's advantage isn't the model. It's the distribution.
- Pre-installed in every GitHub org that pays for it
- Zero friction to enable
- Shows up in the workflow developers already use
- No new vendor to approve, no new tool to learn
- Can bundle pricing into existing subscriptions
CodeRabbit requires you to:
- Find and evaluate it
- Get approval and install it
- Configure and maintain it
That friction compounds. Copilot wins by being there.
The Consolidation Already Happened
Look at this:
Top 3 AI review agents control 72% of all activity.
| Rank | Agent | Share |
|---|---|---|
| 1 | CodeRabbit | ~33% |
| 2 | Copilot | ~29% |
| 3 | Gemini | ~10% |
| Everyone else | ~28% |
And it's getting worse for the long tail.
Korbit - raised money, had traction, purpose-built for code review - shut down this year.
The market consolidated before most people realized there was a market.
The Long Tail Is Crowded
Beyond the top 10, dozens of AI review agents are fighting for what's left:
| Agent | Orgs |
|---|---|
| Cubic AI | ~200 |
| Ellipsis | ~170 |
| OpenHands | ~165 |
| Bolt AI | ~130 |
Based on observable public repo installs
And there are dozens more. All competing for the 28% not owned by the top 3.
The Platform Giants Are Coming
What happens when OpenAI and Google actually start pushing their code review features?
Look at org adoption growth from October to November alone:
| Agent | Oct 2025 | Nov 2025 | Growth |
|---|---|---|---|
| ChatGPT | 4,968 | 7,523 | +51% |
| Gemini | 4,603 | 5,674 | +23% |
ChatGPT is pulling away. The gap between them grew 5x in a single month.
And Gemini still grew 43x this year - from basically nothing to #3 in PR volume globally.
Neither OpenAI nor Google is even trying yet. No big marketing push. No deep IDE integration. No bundling with Workspace or Cloud.
And they have the same platform leverage as Microsoft. Millions already subscribe to ChatGPT and Google Workspace for other use cases. Code review becomes a free add-on, not a new line item.
When they flip that switch, what happens to Sourcery, Greptile, Ellipsis, and the rest of the long tail?
The Adoption Curve Is Insane
Let me show you how fast this moved:
14x in 18 months.
1 in 7 PRs now has an AI reviewer participating. Not a prediction. Already happening.
What This Means If You're Building Dev Tools
I'll be direct:
Distribution > Features. Copilot proved it. Being native to the platform beats being best-in-class at the feature.
The platform tax is real. If GitHub/Microsoft decides your feature is worth building, you're competing against free + pre-installed.
Consolidation is faster than you think. Some players won't survive 2026.
OpenAI and Google are the wildcards. Both growing fast with virtually zero marketing effort for code reviews.
My 2026 Prediction
- Copilot, ChatGPT, and Gemini lock in the top 3. Claude and Cursor climb to #4 and #5.
- AI agents author 2M+ PRs (up from 99K in 2025).
- AI-reviewing-AI becomes a thing - agents reviewing agent-authored code grows 10x.
- At least 2 long-tail players shut down or get acqui-hired.
The Full Data
Everything I cited is from our State of AI Code Review 2025 report:
Includes monthly breakdowns, full agent rankings, and the methodology for how we identified and classified AI agents.
Methodology note: We filtered for active repos only - at least 10 PRs/month and 0.3 feedback events per PR. This filters out noise and surfaces repos with real development activity.
I Want the Counterarguments
Seriously. Tell me where I'm wrong:
- Is "platform beats product" too simplistic?
- Are there code review startups that can survive the consolidation?
- Is the Copilot dominance overstated because of GitHub's visibility bias?
- Am I underestimating how much enterprises care about best-in-class vs. good-enough-and-integrated?
I've been staring at this data for weeks. I want someone to challenge it.
I work on PullFlow - we're agent-agnostic, building unified code review across GitHub, Slack, and AI agents. This research came from trying to understand where the market is actually going.
Top comments (3)
It's also super interesting to see which agents authored pull requests!
I bet we'll see a lot more of this in 2026! 🤯
The speed of consolidation here is the wild part. It feels like the market closed before most teams realized it existed.
The data makes a strong case that “better review” wasn’t enough once Copilot became default. Curious if you think workflow ownership is the only remaining wedge for non-platform tools.
In the early days, yes. There are still many cards left in the deck for non-platform tools: specialization (the best agent for Django code), education (teaching rather than just correcting), automated testing (stronger unit and integration tests), and more.
Any of these can be effective in carving out a niche once the dust settles.