CodeRabbit report finds AI-written code creates 1.7x more bugs

TL;DR

CodeRabbit released a large-scale study on December 17, 2025 comparing 470 AI-generated and human-written pull requests. The report finds AI-authored code produces about 1.7x more issues on average, with higher rates of critical, security, and performance defects. Media coverage highlights that AI-generated code requires more intensive review despite productivity gains.