AI Now Reviews 60% of Bot Pull Requests on GitHub, Tripling from 20% a Year Ago

Summary

AI coding assistants have dramatically increased their presence on GitHub, now reviewing 60% of all bot pull requests—up from approximately 20% in early 2025. This surge was catalyzed by major model releases, particularly Anthropic's Claude 3.7 Sonnet and OpenAI's GPT-4.5 in February 2025, which nearly doubled penetration within weeks. CodeRabbit leads the category with 179,965 PR reviews in the last 30 days, nearly doubling GitHub Copilot's 91,596 reviews.

Beyond code review, AI coding assistants are making inroads into PR creation and code commits. Copilot ranked #4 in PRs opened with 88,943 submissions, while app-building platforms like Lovable Dev demonstrate the broader potential of AI agents writing and submitting code. However, significant untapped opportunity remains in issue creation, where AI accounts for less than 1.5% of bot-created issues, suggesting this remains a frontier requiring deeper project context understanding.

• Issue creation remains AI's weakest category at <1.5% penetration, indicating limitations in automated bug triage and context understanding
• Emerging agentic workflows show teams syncing issues from Linear to GitHub where AI agents autonomously open PRs, hinting at deeper workflow automation

Editorial Opinion

The data reveals both the transformative velocity of modern AI coding tools and their uneven capabilities across the development lifecycle. While PR review automation has become mainstream—a remarkable achievement given near-zero penetration just 18 months ago—the stubborn resistance in issue creation exposes a real gap: understanding user intent, project context, and bug severity remains harder to automate than pattern-matching code. This asymmetry suggests the next frontier isn't just better models, but AI systems that can participate meaningfully in requirements gathering and triage, not just execution.