Google Says AI Now Generates 75% of Its New Code
Google has revealed that artificial intelligence now produces approximately 75% of its new code, marking a significant leap from 25% in 2024 and 50% in 2025.
The company said the AI-generated code is still reviewed by human engineers, but the latest figures highlight how AI has evolved from a supporting tool into a primary driver of software development within the organisation.
This shift is largely powered by Google’s in-house Gemini models, which engineers use for tasks such as generating, refactoring, and migrating code. The company has also expanded AI adoption beyond engineering teams, in some cases linking usage to employee performance evaluations.
Chief Executive Officer Sundar Pichai described the transition as part of a broader move toward what he called “agentic workflows,” where engineers oversee systems capable of executing complex tasks with minimal human input.
According to Pichai, the approach has already delivered measurable productivity gains. He noted that a recent large-scale code migration, carried out through collaboration between engineers and AI agents, was completed six times faster than similar efforts a year earlier.
The development signals a shift away from basic code-assistance tools toward more advanced systems capable of planning and executing entire development processes, including large-scale refactoring and system-wide updates.
However, adoption has not been entirely uniform across the company. Some teams within Google DeepMind have reportedly used external tools such as Claude Code alongside internal systems, leading to internal discussions about standardisation.
Google’s move reflects a broader trend across the tech industry. Microsoft has reported that between 20% and 30% of code in some projects is now AI-generated, with CEO Satya Nadella predicting even higher adoption in the coming years. The company’s CTO, Kevin Scott, has suggested that up to 95% of code could be produced by AI within five years.
Similarly, Meta has set internal targets for AI-assisted coding, while Snap Inc. says at least 65% of its new code is already generated using AI.
The growing reliance on AI is reshaping the role of software developers, with engineers increasingly acting as supervisors of automated systems rather than writing most code manually. At Google, where AI now dominates new code production, that transformation is already well underway.
