75% of Google's New Code Is Now AI-Generated — What That Really Tells Us
Google just dropped a number that will ripple through every engineering organization in the world: 75% of its new code is now generated by AI. Engineers review it, but they're no longer writing most of it from scratch. Up from 50% just last fall, this statistic marks a genuine inflection point in how the world's most sophisticated software company builds software — and it's a preview of where the entire industry is heading.
What's Actually Happening
Google CEO Sundar Pichai disclosed the figure, noting that AI-generated code is reviewed by human engineers before merging. The tooling spans Google's internal AI coding assistants, which are deeply integrated into their development workflows across Search, YouTube, Cloud, and every other major product line.
This isn't Google experimenting with AI for toy projects. This is the core engineering output of one of the most productive software organizations on Earth — and three-quarters of its new lines of code are now coming from models, not humans typing.
Why It Matters
The implications split into two camps. Optimists see this as an enormous productivity multiplier — engineers spending less time on boilerplate and more time on architecture, product thinking, and hard problems. Pessimists see a quality time bomb: AI-generated code tends to be syntactically correct but semantically risky, introducing subtle bugs and security vulnerabilities that pass review because reviewers are moving faster.
For the broader developer economy, the 75% figure is a canary in a coal mine. If Google is here, smaller companies will follow. The question of what "software engineering" means as a profession is no longer hypothetical. The role is evolving — faster than most engineering curricula are adapting. For context on how Google is building the AI infrastructure enabling this shift, see our coverage of Google Workspace Intelligence and its AI-first approach to productivity.
My Take
The number is real, but the framing matters enormously. "75% of new code is AI-generated" is very different from "75% of production logic is AI-written." New code includes tests, scaffolding, boilerplate, documentation stubs — the stuff that's tedious to write but easy to verify. If AI handles those and engineers focus on the hard parts, that's genuinely good engineering leverage.
The risk I'd watch is review quality at scale. When AI generates most of the code and engineers are reviewing it quickly, the social pressure to approve compounds. Bad code gets through not because engineers are lazy, but because the review process wasn't designed for AI-scale output velocity. Google is sophisticated enough to manage this. Most companies aren't.
FAQ
Does this mean Google engineers are being replaced? Not in the near term. Engineers are reviewing, directing, and shipping the AI-generated code — the role is shifting toward oversight and architecture, not disappearing.
Is AI-generated code safe? It can be, when reviewed properly. The concern is that AI code often passes surface-level review but introduces subtle logic errors or security issues that only surface later.
What tools is Google using? Google uses internal AI coding assistants integrated into its development environment, alongside external models. The exact toolchain isn't fully disclosed.
Will other companies reach 75% soon? Many already claim similar figures for specific tasks. Industry-wide, the trend is clear — AI code generation is becoming the default, not the exception.
Related Articles
- Google Workspace Intelligence Wants to Know Everything About How You Work
- SpaceX Has an Option to Buy Cursor for $60 Billion