Google Reports 75% of New Code Now Written by AI

Apr 23, 2026

Google disclosed that 75% of its new code is now AI-generated and human-reviewed, up from 50% reported in February 2025, representing a major shift in software development practices. The company faces internal fragmentation with AI coding capabilities spread across six products including Gemini Code Assist, Jules, and Antigravity, prompting Chief AI Architect Koray Kavukcuoglu to consolidate tools under the Antigravity platform. Despite Google's $2.4-billion Windsurf acquisition and billions invested in Anthropic, some engineers prefer competitor Anthropic's Claude Code, creating tension as most employees are banned from using external tools due to security concerns. The rapid AI adoption has sparked philosophical clashes between AI researchers prioritizing speed and traditional senior engineers maintaining code quality standards, while capacity constraints and performance review pressures complicate implementation across teams at Google Cloud, DeepMind, and Android divisions.

Google's fragmented AI coding product strategy and internal confusion

▪Google's AI coding tools lineup includes Gemini Code Assist, Gemini CLI, AI Studio, Firebase Studio, Jules, and Antigravity.
▪Kathy Korevec, who oversaw Jules at Google, joined OpenAI in early 2025.
▪Google Chief AI Architect Koray Kavukcuoglu is working to unite Google's internal artificial intelligence coding tools under Antigravity in the coming weeks.
▪Google released its Antigravity platform in 2024 following the acquisition of talent and technology from startup Windsurf in a $2.4-billion deal.
▪Google Cloud, Google Core, Google Labs, Android, and DeepMind are all pushing AI coding in different ways at Google.
▪Google appointed Koray Kavukcuoglu to the new position of chief AI architect in 2024 to fold generative AI into Google products.

Competitive pressure from Anthropic and OpenAI in AI coding tools

▪Some teams at Google DeepMind, including those working on the Gemini model, internal applications, and open source models, use Claude Code.
▪Google released Gemini 3 in late 2024, a model that appeared to outperform rival services across a range of benchmarks.
▪Anthropic cut off OpenAI's access to Anthropic's models in 2024.
▪Most Google employees are banned from using competing tools such as Claude Code or Codex due to security concerns.
▪Some Google engineers prefer to use Anthropic's Claude Code over Google's internal AI coding tools.

Internal adoption challenges and cultural tensions at Google

▪John Jumper, who won the Nobel Prize in 2024 alongside Google DeepMind CEO Demis Hassabis, is working on AI coding at Google.
▪Brian Saluzzo, one of the executives who oversaw efforts to promote AI coding within Google, recently left Google.
▪Google DeepMind is forming a new AI coding team led by research engineer Sebastian Borgeaud.
▪Alphabet stated in February 2025 that roughly 50% of new code at Google is written by AI.