GitHub Copilot's grip slips to 51% as Cursor and Claude Code post the fastest IDE debuts on record
Stack Overflow's 2026 Developer Survey shows GitHub Copilot's share among professional developers falling from 67% to 51% in a year, while Cursor and Claude Code debuted at 18% and 10% respectively — the fastest first-year adoption curves the survey has ever recorded.
4 August 2026
Stack Overflow’s 2026 Developer Survey puts a hard number on a shift that’s been visible anecdotally for months: GitHub Copilot’s share among professional developers has fallen from 67% to 51% year over year. In the same survey, Cursor debuted at 18% and Claude Code at 10%, both from a standing start — the fastest first-year IDE debuts the survey has recorded since it started tracking tool adoption. Developers are also spreading their usage: experienced developers now run 2.3 AI coding tools on average, rather than settling on one.
That last point matters more than the headline number. Cursor and Claude Code aren’t cannibalising VS Code, and they’re not purely cannibalising Copilot either — a lot of teams are running Copilot alongside a dedicated AI-native tool for different parts of the job. What’s actually shrinking is the assumption that whichever tool came bundled with your editor is good enough on its own. Eighty-four percent of respondents now use or plan to use AI tools, up from 76% in 2024, so the category is still growing overall even as any single vendor’s share of it gets squeezed.
This is a different survey and a different metric from the JetBrains loyalty data covered here in July — JetBrains measured satisfaction and stickiness (Claude Code’s 91% CSAT), Stack Overflow measures raw usage share across a broader developer population. Both point the same direction: the market that used to be “Copilot, by default” is now genuinely multi-vendor, and newer entrants are winning meaningful share fast rather than staying niche.
So what
If you’re commissioning software and a development partner’s pitch leans on which AI tool they use as a differentiator, this data says that’s weaker ground than it was a year ago — most credible teams now run more than one, and the honest answer to “which tool” is often “it depends on the task.” What’s worth asking instead is how a team decides which tool to reach for, and how they catch the tool’s mistakes before they ship. That’s the standard we hold our own process to — see AI-assisted development, or get in touch to talk through your project.