JetBrains AI Coding Tools Survey 2026: What Developers Actually Use at Work

JetBrains published their AI Pulse survey in January 2026, covering 10,000+ developers worldwide on which AI coding tools they actually use at work — not just awareness, but regular daily usage. The headline finding: 90% of developers use AI tools broadly, but adoption of specialized coding assistants is more concentrated than awareness numbers suggest.

Survey Methodology: JetBrains AI Pulse January 2026 (10,000+ Developers Worldwide)

The JetBrains AI Pulse January 2026 survey polled over 10,000 professional developers across company sizes, industries, and geographies, making it the largest independent snapshot of AI coding tool adoption published in 2026. The survey distinguishes between awareness (have you heard of this tool?), personal use (do you use it for personal projects?), and work adoption (do you regularly use it at your job?) — a three-way distinction that reveals significant gaps between mindshare and real deployment. JetBrains ran parallel surveys in April–June 2025 and September 2025, enabling longitudinal tracking of adoption curves that reveals which tools are accelerating and which are plateauing. The methodology weights responses by developer seniority and company size to prevent startup-heavy or enterprise-heavy skew, giving a representative cross-section of the professional developer population. Key caveats: the sample over-represents JetBrains IDE users (IntelliJ, PyCharm, WebStorm) relative to the broader developer market, which may slightly underweight VS Code-heavy ecosystems where Cursor and GitHub Copilot have stronger native integrations. Despite this, the directional findings are corroborated by multiple independent market research sources and represent the most rigorous published data set on AI coding tool adoption as of early 2026.

Overall AI Adoption: 90% Use AI at Work, But Only 18% Use Specialized Coding Tools

The most striking finding in the JetBrains AI Pulse January 2026 data is the adoption gap: 90% of developers regularly used at least one AI tool at work in January 2026, but only 18% report using specialized agentic coding assistants like Claude Code or Cursor as their primary work tool. The gap is explained by the definition of “AI tool” — ChatGPT for documentation, Gemini for code explanation, and general chatbots dominate broad usage. ChatGPT is used by 28% of developers for coding-related queries, Gemini by 8%, and Claude’s chatbot by 7%. These chatbot use cases are largely additive to existing workflows rather than replacing them. The specialized coding assistant category — tools that operate directly in the IDE or terminal, have agentic file-editing capabilities, and integrate deeply with codebases — is where the 18% threshold represents genuine workflow transformation. This distinction matters for interpreting the market: “90% AI adoption” is accurate for broad usage, but the market for deep coding workflow tools is still in a 15–25% penetration phase that represents significant headroom for continued growth. The next wave of adoption growth will likely come as developers who currently use chatbots for isolated queries shift to agentic tools that handle multi-file workflows end-to-end.

Tool-by-Tool Breakdown: GitHub Copilot, Cursor, Claude Code, Chatbots, and More

GitHub Copilot leads work adoption at 29% with 76% awareness — the largest absolute install base of any specialized coding tool. However, the growth story is unfavorable: Copilot’s adoption has stalled year-over-year with limited movement despite its awareness dominance. Cursor and Claude Code are tied at 18% work adoption, sharing second place by a significant gap from Copilot. Cursor’s strength is its VS Code-native IDE experience with deep agentic capabilities; Claude Code’s strength is its terminal-native workflow and multi-file refactoring quality. JetBrains’ own AI Assistant is used regularly by 9% of developers — a meaningful number, but lower than its IDE-native distribution advantage would suggest, reflecting competition from external tools even within JetBrains’ own user base. Junie, JetBrains’ newer agentic tool, sits at 5% adoption. Google Antigravity reached 6% adoption by January 2026 — notable for a tool that launched mid-cycle, with a 120% growth rate that makes it the fastest new entrant in the market. Among chatbot-style coding helpers, ChatGPT at 28% is by far the dominant general-purpose tool, used heavily for documentation, code explanation, and isolated snippet generation where the overhead of a full coding assistant isn’t warranted.

The Claude Code Surge: From 3% to 18% Work Adoption in Under a Year

Claude Code’s adoption trajectory is the most notable growth story in the JetBrains AI Pulse data. In April–June 2025, Claude Code had approximately 3% work adoption. By September 2025, it reached 12%. By January 2026, it hit 18% — a 6x increase in under a year, and the fastest year-over-year growth rate of any specialized coding tool in the survey. Awareness followed a similarly steep curve: 31% in April–June 2025, rising to 49% in September 2025, then 57% in January 2026. The awareness-to-adoption conversion rate is also the highest in the category, indicating that developers who try Claude Code are more likely to adopt it as a primary work tool than those who try competing products. The product satisfaction metrics explain the retention: Claude Code’s CSAT score of 91% and Net Promoter Score of 54 are the highest in the AI coding tools market, significantly ahead of GitHub Copilot. The 6x growth narrative is especially relevant for startup and individual developer segments — 75% of startups report using Claude Code as their primary coding assistant, compared to 56% of large enterprises (10,000+ employees) who prefer GitHub Copilot. The enterprise-startup split reflects organizational dynamics: enterprise security reviews, IT procurement processes, and existing Microsoft integrations favor Copilot, while startups optimizing for developer velocity gravitate toward Claude Code’s multi-file agentic capabilities.

Enterprise vs Startup Split: Why Company Size Determines Which Tool Teams Choose

The organizational context in which developers work is the strongest predictor of which AI coding tool they use, more than technical preference or individual familiarity. At startups, Claude Code has 75% adoption as the primary coding assistant. At large enterprises with 10,000+ employees, GitHub Copilot dominates at 56%. This split reflects three structural factors: procurement and security review timelines, existing toolchain integration, and the nature of the work being done. Enterprise procurement cycles favor tools with existing enterprise agreements — GitHub Copilot benefits enormously from its integration with GitHub Enterprise and Microsoft 365 subscriptions that IT departments already manage. Security reviews for new AI tools at large enterprises can take 6–18 months, meaning tools that entered the enterprise pipeline earlier have structural advantages regardless of current product quality. Startups face no such procurement friction: a developer can adopt Claude Code in minutes with a personal API key and expense reimbursement. The nature of work also differs: enterprise developers frequently work on legacy codebases with strict compliance requirements, where GitHub Copilot’s GitHub Enterprise audit logging and IP indemnification features matter. Startup developers building greenfield applications care more about multi-file refactoring speed and context window size, where Claude Code’s technical capabilities excel.

Developer Satisfaction Rankings: Claude Code Leads on CSAT (91%) and NPS (54)

Developer satisfaction data from the JetBrains survey reveals a growing quality gap between the market leader by usage (GitHub Copilot at 29% adoption) and the market leader by satisfaction (Claude Code at 91% CSAT and NPS 54). Customer satisfaction score (CSAT) measures the percentage of users who rate their experience positively; Net Promoter Score (NPS) measures the likelihood of recommendation. Claude Code’s 91% CSAT is the highest in the AI coding tools category, followed by Cursor at approximately 85%. GitHub Copilot’s CSAT has declined as its feature velocity slowed relative to newer competitors, and developer expectations for AI coding tools have risen significantly since Copilot launched. The NPS gap is particularly meaningful for predicting future adoption trajectories: high NPS drives organic word-of-mouth adoption, which has been the primary growth engine for both Claude Code and Cursor. GitHub Copilot’s enterprise distribution advantage is real but increasingly offset by the satisfaction gap — developers at large companies who use Copilot because IT mandated it are more likely to advocate for switching to Claude Code than the reverse. Product loyalty metrics are becoming a leading indicator for future market share as the initial adoption wave matures into a replacement cycle.

Productivity Reality Check: Hours Saved vs Code Quality and Review Burden

The productivity data in JetBrains’ companion AI workflow impact report presents a more nuanced picture than headline adoption numbers suggest. Nearly 9-in-10 developers save at least 1 hour per week using AI tools, and 1-in-5 save 8 or more hours per week — evidence of genuine productivity gains that are driving continued adoption. However, 40% of developers with fewer than 10 years of experience report that AI-generated code looks correct on the surface but proves unreliable in testing, and that reviewing AI-generated code takes more effort than writing their own. This productivity paradox — time saved in generation offset by increased review burden — is the central quality challenge facing AI coding tools in 2026. The experience split is significant: senior developers with 10+ years of experience report cleaner net productivity gains, because they can faster identify AI errors and course-correct. Developers earlier in their careers, who lack the pattern-recognition to quickly spot subtle bugs in plausible-looking AI output, experience more review overhead that partially offsets generation speed. The practical implication for teams: deploying AI coding tools without calibrating review processes for AI-generated code leads to underrealized productivity gains and, in some cases, introduces more bugs than it prevents. Teams that report the highest net productivity gains from AI tools also invest in code review practices specifically designed for AI-generated output.

What’s Coming: Google Antigravity, JetBrains Junie, and the Next Wave of Tools

The January 2026 JetBrains data captures a market in rapid transition, with several indicators pointing toward a more competitive landscape by end of 2026. Google Antigravity’s 120% growth rate is the clearest signal of a credible new entrant — reaching 6% adoption from a standing start in under six months demonstrates genuine product-market fit beyond early adopter enthusiasm. Antigravity’s deep Google Cloud integration and first-party Gemini model access position it for strong enterprise adoption among organizations already committed to Google Workspace. JetBrains Junie at 5% is positioned as the in-IDE agentic alternative to terminal-native tools like Claude Code — the IDE-native approach is more accessible for developers who prefer staying in their editor rather than switching to a terminal workflow. The upcoming competitive pressure on GitHub Copilot comes from multiple directions: Claude Code and Cursor from the agentic CLI space, Antigravity from the Google enterprise stack, and Junie from the JetBrains IDE ecosystem. Microsoft’s response will be the critical variable — whether Copilot accelerates its agentic capabilities to match Claude Code’s multi-file reasoning quality, or whether enterprise lock-in is sufficient to maintain market position without feature parity. The satisfaction gap suggests the latter strategy carries long-term risk as developer choice within enterprises grows.

FAQ

The JetBrains AI Pulse January 2026 survey of 10,000+ developers is the most comprehensive published dataset on AI coding tool adoption and provides concrete answers to the questions engineering leaders most frequently ask when evaluating or deploying AI coding tools. The survey’s three-way distinction between awareness, personal use, and work adoption cuts through the noise that inflates headline adoption claims in vendor marketing. The data shows 90% broad AI tool usage, but only 18% specialized coding assistant adoption at work — a gap that represents both the current state and the growth opportunity for the market. The 6x year-over-year growth of Claude Code, GitHub Copilot’s stalling growth despite 76% awareness, and the enterprise-startup divide are the three most actionable findings for tool selection decisions in 2026. These FAQs address the questions most commonly raised when teams evaluate or reconsider their AI coding tool strategy based on this data.

Why has GitHub Copilot’s growth stalled despite being the most recognized AI coding tool?

GitHub Copilot’s 76% awareness and 29% work adoption represent a conversion ceiling that suggests the tool has reached near-maximum penetration among developers predisposed to adopt it. The stalling reflects three factors: feature velocity has slowed relative to Claude Code and Cursor, which ship major capability updates more frequently; developer satisfaction scores have declined as expectations rose faster than Copilot’s capabilities improved; and the tool’s core inline autocomplete experience faces displacement by more powerful agentic tools that handle multi-file tasks. Copilot’s enterprise distribution advantage — built into GitHub Enterprise and Microsoft 365 — maintains its install base, but developers within those organizations increasingly use Claude Code or Cursor alongside Copilot for higher-complexity tasks.

Is the 90% AI tool adoption figure misleading given that only 18% use specialized coding assistants?

The two numbers measure different things. The 90% figure counts developers who use any AI tool for any coding-related purpose — this includes ChatGPT for documentation, Gemini for explaining error messages, and general LLM chatbots for isolated code questions. These are legitimate productivity uses but represent lightweight, additive AI usage rather than deep workflow integration. The 18% figure counts developers using specialized agentic coding assistants (Claude Code, Cursor) as a regular part of their development workflow — file editing, multi-step refactoring, automated PR creation. Both numbers are accurate for their respective definitions. The important takeaway is that broad AI tool usage is near-saturated at 90%, while the market for deep coding workflow transformation at 18% still has significant growth headroom.

What explains Claude Code’s 6x growth from 3% to 18% work adoption in under a year?

Three factors drove Claude Code’s exceptional adoption curve. First, product quality: Claude Code’s multi-file reasoning, context window utilization, and multi-step refactoring capabilities represent a generation ahead of inline autocomplete tools. Second, satisfied users drive referrals: the 91% CSAT and NPS 54 are the highest in the category, meaning most users recommend it actively. Third, startup developer culture: startups adopt tools based on productivity impact without enterprise procurement friction, and the 75% startup adoption rate reflects Claude Code’s product-led growth in the segment with the fastest decision cycles. The awareness curve (31% → 49% → 57% in three survey waves) shows developer attention followed adoption rather than preceding it — the opposite of how marketing-driven tools grow.

Should enterprises switch from GitHub Copilot to Claude Code based on this data?

Not necessarily, and the data does not support a universal recommendation to switch. The survey shows enterprises with 10,000+ employees choose GitHub Copilot at 56% for legitimate structural reasons: Microsoft/GitHub integration, enterprise security agreements, IP indemnification, and procurement simplicity within existing contracts. The satisfaction gap is real, but switching costs in large organizations are also real. A more measured interpretation: enterprises should run controlled pilots of Claude Code or Cursor alongside Copilot for teams doing high-complexity work (multi-file refactoring, large codebase navigation, complex debugging), measure the net productivity delta with their specific workloads, and make migration decisions based on measured outcomes rather than aggregate survey data. The enterprise vs startup split in the survey reflects structural procurement realities as much as product preference.

How should the 40% productivity paradox finding affect AI coding tool deployment strategy?

The finding that 40% of less-experienced developers say reviewing AI-generated code takes more effort than writing their own code does not mean AI tools are counterproductive for those developers — it means the productivity realization depends on how AI output is reviewed. Teams with the highest net AI productivity gains share one practice: they adjust code review processes specifically for AI-generated output, rather than applying the same review standards they use for human-written code. AI-generated code tends to have different failure modes (plausible-looking but logically incorrect logic, missing edge cases, subtle security issues) than human-written code. Review checklists and automated testing suites calibrated to these failure modes recover most of the productivity that unstructured review overhead consumes.