AI Code Assistant Statistics 2026: Copilot, Cursor and Claude Code

Direct Answer

Stack Overflow says 84% of 2025 Developer Survey respondents use or plan to use AI tools in their development process, up from 76% in 2024. JetBrains says 90% of developers regularly used at least one AI coding tool at work in January 2026. GitHub Copilot leads worldwide work adoption at 29%, with Cursor and Claude Code each at 18%.

Cite This Report

Use this URL when citing this report: https://konabayev.com/blog/ai-code-assistant-statistics-2026/. Suggested citation: Konabayev, T. (2026). AI Code Assistant Statistics 2026: Copilot, Cursor and Claude Code. konabayev.com. Last verified May 11, 2026.

Machine-readable copies are available here:

• CSV: /data/ai-code-assistant-statistics-2026.csv
• JSON: /data/ai-code-assistant-statistics-2026.json
• JSONL: /data/ai-code-assistant-statistics-2026.jsonl

Primary source pages used in this report: Stack Overflow 2025 Developer Survey AI section, Stack Overflow 2025 Developer Survey press summary, JetBrains AI coding tools at work research, Microsoft Research GitHub Copilot productivity study, GitHub Copilot Accenture enterprise research, GitHub Octoverse 2025, GitHub Octoverse 2024, DORA trust in AI, DORA 2024 report, and METR arXiv RCT.

Top Citable Claims

These are the most quotable AI code assistant statistics from the audited source set. Do not strip the source domain or caveat when reusing the numbers.

Claim	Figure	Source
Developers using or planning to use AI tools in 2025, up from 76% in 2024	84%	survey.stackoverflow.co
All respondents using AI tools daily in development	47.1%	survey.stackoverflow.co
Professional developers using AI tools daily	50.6%	survey.stackoverflow.co
Developers regularly using an AI tool at work in January 2026	90%	blog.jetbrains.com
GitHub Copilot work adoption worldwide (Jan 2026)	29%	blog.jetbrains.com
Cursor work adoption worldwide (Jan 2026)	18%	blog.jetbrains.com
Claude Code work adoption worldwide (Jan 2026)	18%	blog.jetbrains.com
Developers actively distrusting AI accuracy in 2025, up from 31% in 2024	46%	stackoverflow.co
GitHub Copilot speed gain on a JavaScript HTTP-server task	55.8% faster	microsoft.com
Effect of AI use on completion time in METR open-source RCT	19% longer	arxiv.org
AI-related repositories on GitHub (Octoverse 2025)	Over 4.3 million	github.blog

AI Code Assistant Adoption

AI coding tools moved from early-adopter trend to default workflow during 2024 and 2025. Stack Overflow’s 2025 Developer Survey, with 33,662 responses to the AI section and 26,004 professional developers, says 84% of respondents use or plan to use AI tools in their development process, up from 76% in 2024. Stack Overflow also says 16.2% of all respondents do not plan to use AI tools, leaving a small but persistent holdout.

Usage frequency among all Stack Overflow respondents:

Usage cadence	Share	Source
Daily AI tool use	47.1%	survey.stackoverflow.co
Weekly AI tool use	17.7%	survey.stackoverflow.co
Monthly or infrequent AI tool use	13.7%	survey.stackoverflow.co
Do not plan to use AI tools	16.2%	survey.stackoverflow.co

Professional developers report similar but more intense cadence. Stack Overflow says 50.6% of professional developers use AI tools daily, 17.4% use them weekly, 12.8% use them monthly or infrequently and 14.7% do not plan to use them. By experience, Stack Overflow says 55.5% of early-career developers use AI tools daily and 47.3% of experienced developers use AI tools daily.

JetBrains corroborates the picture with its January 2026 AI Pulse survey of more than 10,000 professional developers worldwide, localized into eight languages. JetBrains says 90% of developers regularly used at least one AI tool at work for coding and development tasks in January 2026, and 74% of developers worldwide had adopted specialized AI tools for developers by the same date. For broader AI tool adoption beyond coding, see our AI marketing tool adoption 2026 report.

Tool Adoption: Copilot, Cursor, Claude Code and Codex

JetBrains AI Pulse (January 2026) is the most current developer survey in this source set for tool-level shares. Brand awareness and adoption at work:

Tool	Aware (Jan 2026)	Used at work (Jan 2026)	Source
GitHub Copilot	76%	29%	blog.jetbrains.com
Cursor	69%	18%	blog.jetbrains.com
Claude Code	57%	18%	blog.jetbrains.com
OpenAI Codex	27%	3%	blog.jetbrains.com
Google Antigravity	Not reported	6%	blog.jetbrains.com

Notes on the tool-level shares:

• JetBrains says GitHub Copilot adoption reaches 40% among developers at companies with more than 5,000 employees.
• JetBrains says Claude Code awareness rose from 31% in April-June 2025 to 49% in September 2025 to 57% in January 2026.
• JetBrains says Claude Code work adoption reached 24% in the US and Canada in January 2026.
• JetBrains says Claude Code had a 91% CSAT and an NPS of 54 in January 2026.
• JetBrains notes that its Codex figure precedes the public launch of the OpenAI Codex desktop app.
• JetBrains says Google Antigravity reached 6% adoption by January 2026 after launching in November.

General-purpose chatbot interfaces remain a meaningful channel for coding work. JetBrains says 28% of developers used the ChatGPT chatbot for coding and development tasks at work in January 2026, while the Gemini chatbot reached 8% and the Claude chatbot reached 7% for the same use case. For more on general-purpose alternatives, see our ChatGPT alternatives guide and our best AI apps roundup.

Treat these figures as developer-survey adoption at work, not revenue share. JetBrains’ research is vendor-published, and “used at work” share does not map directly to billed-seat market share. For cost context on running these tools, see our AI agent costs 2026 breakdown.

GitHub Copilot Productivity and Enterprise Use

Two source types anchor the GitHub Copilot productivity discussion: a controlled task experiment and an enterprise customer study. Microsoft Research ran a controlled experiment where recruited developers implemented an HTTP server in JavaScript with or without GitHub Copilot. Microsoft Research says developers with Copilot completed the task 55.8% faster than the control group. GitHub echoes the headline in its enterprise research with Accenture, saying Copilot helped developers code up to 55% faster in its lab studies.

The GitHub/Accenture randomized controlled trial captures real adoption behavior on top of the speed claim:

Accenture customer research metric	Value	Source
Developers who successfully adopted GitHub Copilot	Over 80%	github.blog
Initial users’ adoption success rate	96%	github.blog
Developers who used Copilot at least five days per week	67%	github.blog
Average Copilot usage per week	3.4 days	github.blog
Developers relying on Copilot in a familiar language	70%	github.blog
Developers receiving and accepting suggestions on install day	96%	github.blog
Average time from first suggestion to first acceptance	1 minute	github.blog
Developers rating Copilot extremely easy to use	43%	github.blog
Developers rating Copilot extremely useful	51%	github.blog
Copilot suggestion acceptance rate	Around 30%	github.blog
Developers reporting they committed Copilot-suggested code	90%	github.blog
Developers reporting team-merged PRs containing Copilot-suggested code	91%	github.blog
Copilot-generated characters retained in the editor	88%	github.blog

GitHub also reports subjective effects from the same study. GitHub says Copilot made 85% of developers feel more confident in their code quality in prior research cited in the Accenture work, that 90% of Accenture developers felt more fulfilled with their job when using Copilot and that 95% of Accenture developers enjoyed coding more with Copilot help.

These results come from GitHub’s own customer research and from a controlled JavaScript task. Both publications encourage readers to treat the figures as customer-specific evidence rather than vendor-neutral market benchmarks. For comparable enterprise SaaS adoption signals, see our SaaS pricing statistics 2026 report.

AI Agents in Development Work

AI agents are now a distinct developer-tooling category, but they are not yet a daily tool for most developers. Stack Overflow asked respondents how often they use AI agents at work:

AI agent usage at work	Share	Source
Daily	14.1%	survey.stackoverflow.co
Weekly	9%	survey.stackoverflow.co
Monthly or infrequently	7.8%	survey.stackoverflow.co
Do not use but plan to	17.4%	survey.stackoverflow.co
Use only copilot or autocomplete, not agents	13.8%	survey.stackoverflow.co
Do not use and do not plan to	37.9%	survey.stackoverflow.co

Among developers who do use AI agents at work, the workload distribution stays heavily coding-centric. Stack Overflow says 83.5% of AI-agent users at work use agents for software engineering, 24.9% use them for data and analytics and 18% use them for IT operations. For the cost side of running agents, see our AI agent costs 2026 breakdown. For the non-developer story on agents in marketing and ops, see what vibe coding means in marketing and our marketing automation statistics 2026 report.

Trust, Accuracy and Human Review

Adoption is high, but trust dropped sharply. Stack Overflow says 46% of developers do not trust the accuracy of AI tool output in 2025, up from 31% in 2024. Stack Overflow says more developers actively distrust AI tool accuracy (46%) than trust it (33%), and only 3% of developers report highly trusting AI tool output.

Frustrations sit on top of distrust. Stack Overflow says 66% of developers cite AI solutions that are almost right, but not quite, as their biggest frustration, and 45% of respondents cite debugging AI-generated code as time-consuming.

When developers do not trust an AI answer, they fall back on people. Stack Overflow asked respondents when they would turn to another person for help instead of an AI tool:

Situation that triggers asking a human	Share	Source
Do not trust the AI answer	75.3%	survey.stackoverflow.co
Have ethical or security concerns about code	61.7%	survey.stackoverflow.co
Want to fully understand something	61.3%	survey.stackoverflow.co
Want to learn best practices	58.1%	survey.stackoverflow.co
Need help fixing complex or unfamiliar code	49.8%	survey.stackoverflow.co

DORA reaches a similar verdict on trust. DORA says 39% of developers outside Google trust the quality of gen AI output only a little or not at all, even while 75% of 2024 DORA respondents outside Google report positive productivity impacts from gen AI. The pattern is consistent across surveys: developers use AI tools heavily, accept productivity gains and still want a human in the loop on accuracy, security and learning.

GitHub Ecosystem and AI Project Growth

GitHub activity tells the same adoption story from the platform side. GitHub Octoverse 2025 covers activity from September 1, 2024 through August 31, 2025:

GitHub Octoverse 2025 metric	Value	Source
AI-related repositories on GitHub, nearly doubling in less than two years	Over 4.3 million	github.blog
Public repositories that used an LLM SDK	More than 1.1 million	github.blog
New public LLM SDK repositories in the previous 12 months, up 178% YoY	693,867	github.blog
Pull requests merged, up 29% YoY	518.7 million	github.blog
New developers using Copilot in their first week	Nearly 80%	github.blog
Issues closed in July 2025	5.5 million	github.blog
Commits made in 2025, up 25% YoY	More than 986 million	github.blog
Monthly code pushes by May 2025	Topped 90 million	github.blog
Share of top 10 open source projects by contributors that were AI-focused	60%	github.blog

The earlier Octoverse 2024 report set the baseline for current AI project growth:

GitHub Octoverse 2024 metric	Value	Source
Open source survey respondents using AI tools for coding or documentation	73%	github.blog
New public and open-source generative AI projects created in 2024	Over 70,000	github.blog
Increase in total contributions to generative AI projects in 2024	Almost 60%	github.blog
Open-source maintainers, verified students and teachers using GitHub Copilot at no cost	More than 1 million	github.blog
Year-over-year growth in complimentary GitHub Copilot users	100%	github.blog
First-time contributors via GitHub Education in the prior year	Over 450,000	github.blog
India year-over-year increase in contributions to generative AI projects	95%	github.blog
France year-over-year increase in contributions to generative AI projects	70%	github.blog

GitHub Octoverse is a platform activity benchmark. GitHub activity and Copilot usage among new GitHub users are not the same as all-developer adoption. For adoption patterns of AI assistants outside developer tooling, see our AI search statistics 2026 report.

Productivity Tradeoffs and Risk Caveats

The productivity story has two careful caveats that engineering leaders should not skip.

First, DORA’s 2024 report finds opposing effects on individuals and on systems. DORA 2024 says AI adoption significantly increases individual productivity, flow and job satisfaction. DORA 2024 also says AI adoption negatively impacts software delivery stability and throughput. The public DORA overview describes the direction for delivery tradeoffs without all effect-size numbers, so cite the direction rather than specific magnitudes.

Second, Stack Overflow gives a mixed self-report on productivity impact. Stack Overflow says 52% of developers agree AI tools or AI agents have had a positive effect on their productivity. On how much AI tools changed development work, Stack Overflow says 16.3% of developers report a great extent of change, 35.3% report some change and 41.4% report not at all or minimally.

Third, a randomized controlled trial published on arXiv by METR pushes back on the universal “AI makes everyone faster” framing. The METR study analyzed 16 experienced open-source developers completing 246 tasks in mature projects where they had an average of five years of prior experience. The allowed-AI condition primarily used Cursor Pro and Claude 3.5 or 3.7 Sonnet.

Key findings from the METR arXiv RCT:

METR RCT finding	Value	Source
Developer forecast of AI impact on completion time (before tasks)	24% shorter	arxiv.org
Developer post-study estimate of AI impact on completion time	20% shorter	arxiv.org
Measured impact of allowing AI on completion time	19% longer	arxiv.org
Economics experts’ prediction prior to study	39% shorter	arxiv.org
ML experts’ prediction prior to study	38% shorter	arxiv.org
Paper length	51 pages, 8 tables, 22 figures	arxiv.org

Treat the METR result as a strong negative signal for that specific setting (experienced maintainers in mature open-source repositories) rather than a universal AI coding result. Pair it with the GitHub/Accenture and Microsoft Research enterprise findings to get a balanced view.

Methodology and Source Notes

This report uses only source-locked numeric claims collected on May 11, 2026, with Firecrawl snapshots saved at collection time. Methodology and caveats by source:

• Stack Overflow 2025 Developer Survey (survey.stackoverflow.co): the AI section drew 33,662 responses, including 26,004 professional developers. The response base changes per question, so use the original survey when making per-subgroup comparisons.
• Stack Overflow 2025 press summary (stackoverflow.co): year-over-year AI adoption, trust and frustration findings at a headline level. Prefer the survey page for response-base detail.
• JetBrains AI Pulse (blog.jetbrains.com): based on the January 2026 AI Pulse survey of more than 10,000 professional developers worldwide, localized into eight languages, plus September 2025 AI Pulse and Developer Ecosystem survey waves. Vendor-published; “used at work” share is not revenue share.
• Microsoft Research (microsoft.com): controlled experiment where recruited developers implemented an HTTP server in JavaScript with or without GitHub Copilot. Strong for the single task; not a generalizable production benchmark.
• GitHub/Accenture customer research (github.blog): randomized controlled trial and adoption analysis with Accenture developers, combining DevOps telemetry and survey responses. Vendor-published customer research.
• GitHub Octoverse 2025 (github.blog): activity on GitHub from September 1, 2024 through August 31, 2025. Platform-level metrics; not all-developer adoption.
• GitHub Octoverse 2024 (github.blog): platform and open source survey context.
• DORA insight article and 2024 report (dora.dev): 2024 DORA survey findings and Google internal research on gen AI, productivity and trust. Treat as directional for delivery tradeoffs.
• METR arXiv RCT (arxiv.org): randomized controlled trial of 16 experienced open-source developers completing 246 tasks. Strong negative signal for the studied setting, not a universal AI coding result.

FAQ

How many developers use AI code assistants in 2026? Stack Overflow says 84% of 2025 Developer Survey respondents use or plan to use AI tools in their development process, up from 76% in 2024. JetBrains says 90% of developers regularly used at least one AI coding tool at work in January 2026, and 74% had adopted specialized AI tools for developers by the same date.

How many developers use AI tools daily? Stack Overflow says 47.1% of all respondents use AI tools daily in their development process, with 17.7% using them weekly and 13.7% using them monthly or infrequently. Among professional developers, 50.6% use AI tools daily. By experience, 55.5% of early-career developers and 47.3% of experienced developers use AI tools daily.

What is the work adoption share of GitHub Copilot, Cursor and Claude Code? JetBrains says 29% of developers worldwide used GitHub Copilot at work in January 2026, with 18% using Cursor and 18% using Claude Code. JetBrains says GitHub Copilot adoption reaches 40% at companies with more than 5,000 employees, and Claude Code work adoption reaches 24% in the US and Canada. Claude Code had a 91% CSAT and an NPS of 54 in January 2026.

Do developers trust AI code assistants? No, trust dropped sharply. Stack Overflow says 46% of developers do not trust the accuracy of AI tool output in 2025, up from 31% in 2024. More developers actively distrust AI tool accuracy (46%) than trust it (33%), and only 3% of developers report highly trusting AI tool output. DORA adds that 39% of developers outside Google trust the quality of gen AI output only a little or not at all.

How much faster does GitHub Copilot make developers? Microsoft Research says developers with GitHub Copilot completed a JavaScript HTTP-server task 55.8% faster than the control group. GitHub says Copilot helped developers code up to 55% faster in its lab studies. In the Accenture customer research, developers accepted around 30% of Copilot suggestions, 90% reported committing Copilot-suggested code and 91% reported teams merged pull requests containing Copilot-suggested code.

How widely are AI agents used in development work? Stack Overflow says 14.1% of respondents use AI agents at work daily, 9% use them weekly and 7.8% use them monthly or infrequently. Another 17.4% do not use AI agents at work but plan to, while 37.9% do not use them and do not plan to. Among AI-agent users, 83.5% use agents for software engineering, 24.9% use them for data and analytics and 18% use them for IT operations.

Does AI always make developers faster? No. DORA 2024 says AI adoption increases individual productivity, flow and job satisfaction but negatively impacts software delivery stability and throughput. A METR arXiv randomized controlled trial of 16 experienced open-source developers on 246 tasks found that allowing AI tools increased completion time by 19%, even though developers forecast a 24% improvement before the study and estimated a 20% improvement after it. Economics and ML experts had predicted 39% and 38% shorter completion times.

How fast is the AI coding ecosystem on GitHub growing? GitHub Octoverse 2025 says AI-related repositories on GitHub exceeded 4.3 million, nearly doubling in less than two years. More than 1.1 million public repositories used an LLM SDK, and 693,867 such repositories were created in the previous 12 months, up 178% year over year. Developers merged 518.7 million pull requests, up 29% year over year, and 60% of the top 10 open source projects by contributors were AI-focused. GitHub Octoverse 2024 adds that 73% of open source survey respondents used AI tools for coding or documentation.

Last verified: May 11, 2026.