The AI-Generated Code Majority Approaches
JetBrains' annual developer survey, covering 12,000 developers across 60 countries, reports that at companies with formal AI coding tool adoption programs, AI-generated or AI-heavily-assisted code now accounts for 45% of new code commits. At companies without formal programs, the figure is 18% as individual developers use AI tools independently.
Tool Dominance
GitHub Copilot remains the market leader by usage share at 54% of AI coding tool users. Cursor IDE (which embeds Claude as its core AI) has rapidly climbed to 23% adoption, with particularly strong representation among individual developers and startups. JetBrains AI Assistant (now integrated across all JetBrains IDEs) is used by 31% of JetBrains IDE users.
Task Distribution
The survey breaks down what developers use AI coding tools for: code completion and generation (used by 91% of AI tool users), documentation writing (67%), unit test generation (61%), code review and explanation (58%), debugging assistance (54%), and architecture design discussion (39%). The last category — AI as an architectural thinking partner — is the fastest-growing use case.
Quality Outcomes
Developer self-reported quality of AI-assisted code versus non-AI code shows complexity: 71% report faster initial implementation, but 52% report AI-generated code requires more careful review than human-written code for security and edge case handling. The consensus finding is that AI tools accelerate experienced developers substantially while potentially creating false confidence in less experienced developers.
Business Productivity Impact
Companies that have implemented formal AI coding programs report 28% faster feature delivery on average, though the methodology for such measurements varies widely across companies.