AI is generating most of Google’s code

AI code reviewed by human engineers makes up three-quarters of the code created inside Google.

According to Business Insider, this number has been steadily climbing. In October 2024, Google said about 25 per cent of its code was AI-generated. That number jumped to 50 per cent last fall.

The company has been increasingly encouraging employees to use AI for both coding and non-coding related tasks. Google CEO Sundar Pichai wrote in a post on Wednesday that Google was reaching peak performance in its workflows.

“Recently, a particularly complex code migration done by agents and engineers working together was completed six times faster than was possible a year ago with engineers alone,” Pichai said.

The engineers are reportedly using Gemini’s models for generating code, with some models specifically having set specific AI goals which will be factored into performance reviews. Some Google DeepMind employees have been permitted to use Anthropic’s Claude AI in the last few months, resulting in some tension between employees.

Google is not the only company pushing to move more of its code to AI.

Microsoft CEO Satya Nadella reportedly said last year that between 20 and 30 per cent of code was generated by AI. The company’s CTO, Kevin Scott, said that he believed that 95 per cent of code could be written by AI by 2030.

Meta is also going in this direction. As of Q4 2025, the goal was to have 55 per cent of code changes from software engineers in some of Meta’s orgs be “Agent-Assisted.” So far in 2026, 65 per cent of engineers in its creation org are expected to write 75 per cent of its code using AI.

How effective the “Agent-Assisting” is definitely up for debate, but only time, and bug reports, will tell how effective the AI direction is for these companies.