How automation peaked at 62% in 2022 and what the decline to 34% means
Peak Bot PRs
in 2022
-28pts
decline by 2024
Based on 294 million PRs analyzed | GitHub Archive / BigQuery | 2020-2024
From 2020 to 2022, bot-generated pull requests steadily climbed to dominate GitHub activity. Then something changed. By 2024, bot PRs had dropped by nearly half.
| Year | Total PRs | Bot % | YoY Change |
|---|---|---|---|
| 2020 | 47.0M | 54.4% | - |
| 2021 | 58.1M | 57.1% | +2.7pts |
| 2022Peak | 76.0M | 62% | +4.9pts |
| 2023 | 56.3M | 40.4% | -21.6pts |
| 2024 | 57.2M | 34.3% | -6.1pts |
The early 2020s saw an explosion in automated pull requests. Several factors drove this growth:
GitHub acquired Dependabot in 2019 and integrated it directly into the platform. By 2022, it was the default for security updates.
GitHub Actions launched in 2019, enabling more teams to automate their workflows and generate PRs programmatically.
Renovate Bot emerged as a powerful Dependabot alternative, offering more configuration options and monorepo support.
Log4j and other high-profile vulnerabilities in 2021 pushed teams to adopt automated dependency updates more aggressively.
"Bot PRs peaked at 62% in 2022, then declined to 34% in 2024. The automation gold rush is over."
The drop from 62% to 34% happened fast—almost halving in just two years. What caused this reversal?
Teams grew tired of noisy Dependabot PRs. Many configured less aggressive update schedules or switched to grouped updates, reducing PR volume.
Rather than merging each dependency update individually, teams adopted batch update strategies—one PR for many changes instead of many PRs.
AI coding assistants (Copilot, etc.) boosted human productivity. Developers shipped more code faster, shifting the ratio back toward human-authored PRs.
In our most recent analysis period, we measured the split in real-time:
3.0M
Human PRs (62.1%)
1.8M
Bot PRs (37.9%)
The bot decline isn't a failure of automation—it's a maturation.
Early Dependabot adoption created PR noise that teams eventually learned to manage. The shift to grouped updates, smarter scheduling, and AI-assisted human coding represents a more sustainable equilibrium. Bots haven't gone away—they've just gotten quieter and more efficient. The 34% that remain are doing more targeted, higher-value work.
"In December 2024, human developers submitted 3 million PRs while bots contributed 1.8 million. The ratio has shifted decisively back to humans."
If you're still drowning in bot PRs, you might be behind the curve. Here's what modern teams are doing:
This analysis covers merged pull requests from GitHub Archive / BigQuery spanning 2020-2024. Bot detection uses login name patterns (containing "[bot]", ending in "-bot", or matching known automation accounts like "dependabot", "renovate", "github-actions"). Year-over-year data represents calendar year totals. December 2024 data represents a single-month snapshot. For full methodology, see the complete study.
CodePulse helps you understand your team's automation patterns and developer activity.