The AI coding tool numbers are in: 90% of developers use one at work, and Claude Code grew 6x in nine months
JetBrains' Developer Ecosystem survey of over 10,000 professional developers puts GitHub Copilot at 29% work usage, Cursor at 18%, and Claude Code at 18% after 6x growth since April 2025 — with Claude Code posting the highest loyalty scores (91% CSAT, NPS of 54) of any tool measured.
17 July 2026
JetBrains’ Developer Ecosystem research, drawn from its September 2025 and January 2026 AI Pulse waves across more than 10,000 professional developers, has put hard numbers on a shift that’s mostly been discussed in vague terms until now. Ninety percent of developers regularly use at least one AI tool at work, and 74% have adopted a tool built specifically for developers rather than a general chatbot. On the specific tools: GitHub Copilot leads work usage at 29%, having largely plateaued; Cursor sits at 18%, also slowing; and Claude Code has grown roughly six-fold since April 2025 to reach 18% as well — effectively level with Cursor and closing on Copilot in under a year.
The more telling number sits underneath the usage share: loyalty. Claude Code posts 91% customer satisfaction and a net promoter score of 54, both the highest of any coding tool JetBrains measured — meaning developers who try it are unusually likely to stick with it and recommend it, not just list it as one of several tools in rotation. In the US and Canada specifically, Claude Code’s work-usage share reaches 24%.
So what
Adoption numbers tell you what’s mainstream; loyalty numbers tell you what’s actually working for the people using it day to day. A team that’s still choosing its primary AI coding tool, or auditing what a development partner uses, now has real data instead of vendor claims to weigh that decision against — and satisfaction scores this lopsided are a stronger signal than market share alone. We use Claude Code as part of our day-to-day delivery process for exactly the reasons this data shows up in the numbers: reliability and depth of output the team can trust more, not just adopt more. More on how that fits into our build process on AI-assisted development, or get in touch to talk through your project.