Google says AI writes 75% of its new code; independent studies find much lower shares
Google's 75% counts AI-suggested code that engineers approved. Independent studies, using stricter definitions, find much lower shares.
What companies say. In October 2024, Google’s chief executive Sundar Pichai said more than a quarter of new code at Google was generated by AI, then reviewed and accepted by engineers. In April 2026 he put the figure at 75%, up from 50% the previous autumn. Microsoft’s Satya Nadella estimated in April 2025 that 20–30% of the code in Microsoft’s repositories was written by software.
What independent studies find. A study in Science (January 2026) used a classifier, a program trained to spot AI-written code, on more than 30 million Python contributions from about 160,000 GitHub developers. It estimated that the AI-assisted share of US Python functions rose from about 5% in 2022 to about 29–30% by early 2025. Only experienced developers showed a productivity gain, a modest 3.6% more commits. Among the most-depended-on open-source packages, commits (saved code changes) that openly credit AI rose from 0.5% to 5.3% a month between August 2025 and July 2026.
What it means
A growing share of new code is written with AI help, but how big it is depends on who counts and what counts. Company figures include accepted autocomplete suggestions and cannot be checked; independent studies use narrower measures and older data. None of them says how much of the actual engineering work AI does.
Caveats
- Google and Microsoft don’t publish a method; “AI-generated” includes accepted autocomplete suggestions and is hard to fact-check.
- The Google figure is about new code, not the whole codebase.
- Openly credited AI use undercounts actual use; the Science estimate covers only Python and ends in early 2025.
- Share of code is not the same as share of engineering value.