Google has announced a significant acceleration in its adoption of artificial intelligence for software development, with 75% of all new code generated internally now being AI-written and approved by human engineers. This figure marks a substantial increase from approximately 25% in October 2024 and 50% in late 2025, demonstrating a rapid transformation in the company's engineering practices over roughly two years.
The remarkable shift was highlighted by a social media post from "Deedy," stating, > "In ~2yrs, Google has gone from 0% code written by AI to 75% code written by AI. What an incredible transformation of how software is created." This sentiment echoes statements from Google's leadership. CEO Sundar Pichai previously indicated that the company is leveraging AI internally to boost productivity and efficiency, emphasizing that AI-generated code is reviewed and accepted by engineers.
This internal AI adoption is powered by tools such as AI Studio, Gemini CLI, and Jules, which facilitate agentic iteration and seamless integration into existing code review workflows. Pichai has consistently framed AI coding as a productivity multiplier rather than a replacement for human engineers, suggesting that increased efficiency expands the scope of what's possible and thus increases the demand for talent. Google plans to continue hiring engineers, focusing on higher productivity leading to broader innovation.
The rapid integration of AI into Google's coding processes reflects a broader industry trend. Microsoft CEO Satya Nadella reported in April 2025 that around 30% of Microsoft's code was AI-written, with predictions from CTO Kevin Scott that 95% of code could be AI-generated within five years. Anthropic is also noted for having nearly 100% of its code written using AI, underscoring the competitive landscape in AI-driven development.
While the primary focus is on increased velocity and expanded capabilities, the cultural shift within Google has not been without discussion. An internal policy temporarily restricted the use of Google's own Gemini model for internal coding, a decision later reversed. The emphasis remains on the approval rate of AI-generated code by human engineers, ensuring quality and adherence to standards before deployment.
