JetBrains asked more than 15,000 professional developers what they actually use to write code, and the answer moved fast between January and July. That's the real story behind AI coding tool adoption in 2026: not one clear winner settling in, but a sharp reshuffle inside two quarters. Claude Code usage rose from 18% in January to 39% globally by mid-2026, while GitHub Copilot and Cursor both lost ground over the same window. If your team picked a default coding assistant in Q1, that pick was made on data that's now two quarters old.
What JetBrains actually found
The JetBrains Developer Ecosystem Survey 2026 is the tenth edition of an annual survey JetBrains has run since 2016, published August 19, 2026. This wave drew more than 15,000 professional developers worldwide, fielded from May to July 2026 and localized into 8 languages. JetBrains sampled by quota, a fixed number of responses per region, then reweighted the results by region, employment status, programming language, and existing familiarity with JetBrains products. That methodology detail is stated on the page, and most coverage of this release left it out entirely.
Among those 15,000+ respondents, 90% said they use an AI coding agent at least weekly, and 68% use one daily. That's not a niche habit anymore. It's the baseline a renewal decision has to start from.
What moved, and how fast
Four tools have enough history in this survey to show a real trend line. Three don't, because JetBrains didn't publish a January baseline for OpenCode, Google Antigravity or JetBrains AI. Here's what JetBrains measured between the January 2026 baseline and the May-July 2026 fieldwork:
| Tool | Jan 2026 share | May-July 2026 share | Delta |
|---|---|---|---|
| Claude Code | [18%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | [39%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | +21 pts |
| GitHub Copilot | [29%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | [21%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | -8 pts |
| Cursor | [18%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | [12%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | -6 pts |
| Codex | [3%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | [16%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | +13 pts |
| OpenCode | n/a | [7%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | n/a |
| Google Antigravity | n/a | [6%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | n/a |
| JetBrains AI | n/a | [~9%](https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/) | n/a |
Claude Code in the US ran even higher, at 47%. And it didn't just pick up casual, occasional users. JetBrains found that 31% of developers now name it their primary tool, and close to 80% of the people who use it regularly at work name it their single most used tool. That conversion rate is the number that should worry any team that assumed developers would try the new thing and drift back to whatever they already knew.
Why a vendor survey about AI tools deserves one caveat
JetBrains sells JetBrains AI, which shows up in this same survey at roughly 9% adoption. That's a real conflict worth naming directly: a company competing in the market it's measuring published the data describing how that market moved. Add to that the fact every figure here is self-reported, developers stating which tool they use rather than usage logs JetBrains pulled itself, and you've got a genuine limit on how much weight any single number from this survey should carry alone.
None of that is a reason to throw the data out. JetBrains discloses its sampling and weighting methodology on the page, the sample size is large enough to move past noise, and the direction of the shift tracks with everything else reported in the same release. Treat it as one credible, imperfect data point rather than the final word.
What this means for a renewal decision
If your engineering org picked a default AI coding tool in January or February 2026, that decision was made against a market that looked very different than it does now. GitHub Copilot held 29% adoption then; it holds 21% now. Cursor held 18%; it holds 12% now. A seat contract signed in Q1 and renewed on autopilot in Q3 or Q4 is pricing a tool against a market it isn't actually competing in anymore.
That's not an argument to switch to whichever tool JetBrains says is winning this quarter. It's an argument to look before you renew, not after. Pull your own usage data: which tool do engineers actually open most, which one gets abandoned mid-task, which one shows up in code review comments as "the AI wrote this weird." Compare that against what the market did in the same window. If your internal numbers and JetBrains' numbers point the same direction, the renewal is easy to sign. If they don't, you've got a real question to answer first.
A single-default policy also gets harder to justify the faster the market moves. Standardizing on one tool made sense when switching costs were the main friction and the options were roughly interchangeable. With this much churn in six months, a mixed-tool policy, where different teams or individuals run more than one assistant and the org actually tracks which one earns its seat, costs less in flexibility than a wrong single bet costs in lock-in. Tool choice isn't only a productivity question either. It changes how fast a new hire gets productive with whichever tool your team standardizes on, and it should factor into hiring for engineers who can evaluate and review AI-generated code, not just use it. If you're also hiring while this shakes out, that means assessing how a candidate actually uses AI tooling during technical evaluation, not just checking a box for "has used Copilot."
What the survey can't tell you
Market share isn't the same question as whether a tool makes your team faster. JetBrains measured which tool developers picked, not what that tool did to their output. Whether a tool actually ships faster code is a separate question from which tool people picked, and the honest answer to that one depends on the task, not on a tool's market position.
This is also one wave of data set against one earlier baseline, January versus May-July. Two points make a line, not a trend. JetBrains can't tell you whether Claude Code's growth keeps compounding through the rest of 2026 or plateaus the way Cursor's did after its own early run. And there's nothing here on cost per seat or on what any of this shift actually did to delivery outcomes at the team level, only on which tool people said they reach for.
FAQ
How big was JetBrains' 2026 Developer Ecosystem Survey?
More than 15,000 professional developers worldwide, fielded from May to July 2026. It's the tenth edition of the survey, localized into 8 languages, published by JetBrains on August 19, 2026.
How much did Claude Code adoption grow in 2026?
JetBrains reported Claude Code usage rising from 18% in January 2026 to 39% globally by mid-2026, and 47% in the US specifically. 31% of developers now name it their primary tool.
Did GitHub Copilot and Cursor lose share in 2026?
Yes, according to the same survey. GitHub Copilot fell from 29% to 21% and Cursor fell from 18% to 12% over the same January-to-mid-2026 window, while Codex grew from 3% to 16%.
Should we switch AI coding tools based on this survey alone?
No. One survey, even a large one, tells you what the market did, not what your team's output looks like on your actual codebase. Use it as a trigger to check your own usage data before a renewal, not as a mandate to switch tools on its own.