JetBrains: Claude Code Grew 6x in Six Months, Tops All Tools in Satisfaction

JetBrains published two significant pieces of AI coding research this week. One is a 10,000-developer survey tracking which tools developers actually use at work. The other is a peer-reviewed behavioral study analyzing 151 million logged events from 800 developers over two years. Together they offer the clearest picture yet of how the AI coding tool market is actually developing.

The Adoption Numbers

The survey ran in January 2026 across 10,000-plus professional developers in eight languages, with a follow-up wave compared to September 2025 data. The headline finding: 90% of developers regularly used at least one AI tool at work for coding.

GitHub Copilot leads on workplace adoption at 29%, with the highest awareness of any tool (76% of respondents knew it). But JetBrains notes that Copilot’s growth has stalled, likely a consequence of the pause on new Pro signups and the saturation of its original enterprise beachhead.

Claude Code and Cursor are tied at 18% workplace adoption each. The Claude Code number is the one worth sitting with: the tool was at roughly 3% adoption in April through June 2025. By January 2026, it was at 18%. That’s six times the user base in about six months, a trajectory that doesn’t happen through organic word-of-mouth alone. Anthropic’s decision to bundle Claude Code access into existing Claude subscriptions and heavily promote it to developers accelerated adoption in the back half of 2025.

Claude Code also leads on satisfaction. It has a 91% CSAT score and a Net Promoter Score of 54, both the highest of any tool in the survey. The satisfaction gap between Claude Code users and Copilot users is meaningful, though Copilot’s larger installed base means it still dominates in raw headcount.

Other tools in the data:

• Google Antigravity: 6% workplace adoption despite only launching in November 2025
• JetBrains AI Assistant (or combined with Junie): 9-11%
• OpenAI Codex: 3%, though JetBrains notes this data predates the public desktop launch

The 3% for Codex is worth watching in the next survey wave. The desktop launch in March and the enterprise GSI partnerships announced in April came after this data was collected.

What the Behavioral Data Shows

The companion study, presented at ICSE 2026 in Rio de Janeiro, took a different approach from the survey. JetBrains’ Human-AI Experience (HAX) team analyzed telemetry from 800 developers across two years: 400 who used JetBrains AI Assistant regularly from April through October 2024, and 400 who didn’t touch it during that period. The dataset covered 151,904,543 logged events.

The headline finding cuts against the standard productivity narrative. AI users typed roughly 600 more characters per month than non-users, and that gap grew over time. They also deleted about 100 more lines per month, compared to 7 for non-users. Context switching increased for AI users by six IDE activations per month while non-users actually decreased by seven.

When developers in the survey portion were asked whether AI tools changed their editing habits, most said no. The behavioral logs show something different.

The researchers describe this as AI “redistributing and reshaping developers’ workflows in ways that often elude their own perceptions.” Developers feel more productive and report that AI keeps them in flow. The data shows more editing activity, more deletion, and more context switching. Whether that’s actually better is a different question from whether it feels better.

The one productivity signal that didn’t change: debugging frequency. AI tools didn’t meaningfully reduce how often developers had to diagnose and fix problems, despite that being one of the common claims made for them.

What to Make of This

The adoption data reflects a market moving fast but not uniformly. Copilot has scale; Claude Code has momentum and the highest satisfaction score in the survey; Cursor has held its ground against both.

The behavioral research is harder to spin either direction. It doesn’t show AI tools making developers worse, but it doesn’t validate the “stay in flow” framing that most tools use in their marketing either. What it shows is that developers are doing more, editing more, and switching contexts more often than they report doing. Whether that’s good productivity depends on what’s in those extra characters and deletions.

Both papers are worth reading in full. The survey methodology includes geographic weighting and experience-level raking, which makes the adoption numbers more reliable than most self-reported industry surveys. The behavioral study’s two-year longitudinal design is genuinely unusual for this kind of research.

---

Sources: JetBrains AI Coding Tools Adoption Survey (2026), JetBrains HAX Behavioral Study, ICSE 2026