Google CEO Sundar Pichai says the once-popular question of what percentage of Google’s code is written by AI no longer makes sense to ask, because AI now touches virtually every line of code the company ships.
Pichai has previously cited a specific figure for how much of Google’s code was AI-generated — he had said that number was around 25 percent, a figure that later climbed past 30 percent. But asked in a recent interview what that number stands at today, Pichai said the metric itself has stopped being a meaningful way to measure AI’s role in how Google builds software.
“That number in our world today is meaningless in the sense that every part of code we now touch is influenced by AI deeply,” Pichai said.
Instead of a single percentage, Pichai pointed to what Google is seeing from Argon, the coding agent it has rolled out internally, as a better signal of how deeply AI has embedded itself into engineering workflows. He said Google is now able to track how engineers are using the tool to take on far more ambitious work than simple code completion.
“What we are tracking with Argon, for example, we are now able to see it’s been a point of inflection. People are using Argon to really go tackle complex projects, including projects which affect millions of lines of code,” he said.
Pichai backed that up with numbers from the weeks since Argon’s rollout. “In the four weeks since we have launched Argon, we have seen a 30 percent increase in CLs, and more importantly, almost a similar increase — one third decrease in rollbacks,” he said, referring to changelists, Google’s internal term for code submissions, and to the rate at which those changes had to be reversed after being pushed. For Pichai, that combination — more code shipped, with fewer of those changes getting rolled back — is the real evidence that AI is making engineers more productive without making their code worse.
“So I think people are writing code more productively and better code with it,” Pichai said. “Nobody would want to code the way they were coding three years ago or even the same time last year.”
The comments mark a shift in how Google’s leadership talks about AI-assisted coding. For much of the past two years, the company has leaned on a single headline statistic — first 25 percent, then north of 30 percent — to describe AI’s footprint in its codebase. Pichai’s latest remarks suggest Google now sees that framing as outdated, with AI less a discrete tool used for a fraction of tasks and more an ambient layer across the entire development process, from small fixes to large-scale migrations spanning millions of lines of code.