AI Coding Statistics 2026: Developer Adoption, Tools, Trust and Productivity
Contents
AI coding tools are widely used, but adoption is outpacing trust—and measured productivity depends heavily on task type, developer experience and workflow context.
Evidence cutoff: September 9, 2026. This 2026 edition uses the newest primary research available at publication. Any later-year figures are identified as forecasts rather than observed results.
Statistics in summary
- 84% used or planned to use AI tools in development, up from 76% in 2024.
- 51% of professional developers used AI tools daily.
- 60% expressed favorable sentiment toward AI tools, down from more than 70% in 2023 and 2024.
- 61% vs. 53% of professionals versus learners expressed favorable sentiment.
- 46% distrusted the accuracy of AI-tool output.
- 33% trusted AI-tool accuracy.
- 3% highly trusted AI-tool output.
- 20% of experienced developers highly distrusted AI output.
- 64% did not see AI as a threat to their jobs, down from 68% in 2024.
- 31% currently used AI agents.
- 17% planned to use AI agents.
- 38% had no plans to use AI agents.
Adoption and sentiment statistics
- 84% used or planned to use AI tools in development, up from 76% in 2024. (Stack Overflow, 2025)
- 51% of professional developers used AI tools daily. (Stack Overflow, 2025)
- 60% expressed favorable sentiment toward AI tools, down from more than 70% in 2023 and 2024. (Stack Overflow, 2025)
- 61% vs. 53% of professionals versus learners expressed favorable sentiment. (Stack Overflow, 2025)
- 46% distrusted the accuracy of AI-tool output. (Stack Overflow, 2025)
- 33% trusted AI-tool accuracy. (Stack Overflow, 2025)
- 3% highly trusted AI-tool output. (Stack Overflow, 2025)
- 20% of experienced developers highly distrusted AI output. (Stack Overflow, 2025)
- 64% did not see AI as a threat to their jobs, down from 68% in 2024. (Stack Overflow, 2025)
Agents and productivity statistics
- 31% currently used AI agents. (Stack Overflow, 2025)
- 17% planned to use AI agents. (Stack Overflow, 2025)
- 38% had no plans to use AI agents. (Stack Overflow, 2025)
- 69% of developers who used agents at work reported increased productivity. (Stack Overflow, 2025)
- 52% either did not use agents or stayed with simpler AI tools. (Stack Overflow, 2025)
Developer friction statistics
- 66% were frustrated by AI solutions that were almost right but not quite. (Stack Overflow, 2025)
- 45% said debugging AI-generated code was more time-consuming. (Stack Overflow, 2025)
- 75.3% of respondents in the cited user subgroup did not trust AI answers. (Stack Overflow, 2025)
- 61.7% reported ethical or security concerns about code. (Stack Overflow, 2025)
- 61.3% wanted to understand their code fully. (Stack Overflow, 2025)
- 35% visited Stack Overflow because of AI-related issues at least some of the time. (Stack Overflow, 2025)
- 36% learned AI-enabled tools for work or career advancement. (Stack Overflow, 2025)
- 44% used AI tools to learn coding, up from 37%. (Stack Overflow, 2025)
Models and environments statistics
- 81% of developers using LLMs reported OpenAI GPT models. (Stack Overflow, 2025)
- 43% reported Claude Sonnet models. (Stack Overflow, 2025)
- 35% reported Gemini Flash models. (Stack Overflow, 2025)
- 18% used Cursor. (Stack Overflow, 2025)
- 10% used Claude Code. (Stack Overflow, 2025)
- 5% used Windsurf. (Stack Overflow, 2025)
Workplace tool adoption statistics
- 90% regularly used at least one AI tool at work for coding or development tasks. (JetBrains Research, 2026)
- 74% used a specialized AI developer tool rather than only a general chatbot. (JetBrains Research, 2026)
- 76% had heard of GitHub Copilot. (JetBrains Research, 2026)
- 29% used GitHub Copilot at work. (JetBrains Research, 2026)
- 40% of developers at companies with more than 5,000 employees used GitHub Copilot. (JetBrains Research, 2026)
- 69% were aware of Cursor. (JetBrains Research, 2026)
- 18% used Cursor at work. (JetBrains Research, 2026)
- 57% had heard of Claude Code in January 2026, up from 49% in September 2025 and 31% in mid-2025. (JetBrains Research, 2026)
- 18% used Claude Code at work worldwide. (JetBrains Research, 2026)
- 24% used Claude Code at work in the United States and Canada. (JetBrains Research, 2026)
- 91% CSAT was Claude Code’s reported satisfaction score. (JetBrains Research, 2026)
- 54 NPS was Claude Code’s reported likelihood-to-recommend score. (JetBrains Research, 2026)
- 3% used OpenAI Codex at work before the desktop app’s public launch. (JetBrains Research, 2026)
- 6% used Google Antigravity by January 2026. (JetBrains Research, 2026)
- 28% used the ChatGPT interface for coding tasks at work. (JetBrains Research, 2026)
- 11% used JetBrains AI Assistant and/or Junie. (JetBrains Research, 2026)
Randomized productivity evidence statistics
- 16 developers participated in the randomized controlled trial. (Model Evaluation & Threat Research, 2025)
- 246 tasks were completed in mature open-source projects. (Model Evaluation & Threat Research, 2025)
- Five years was participants’ average prior experience with their repository. (Model Evaluation & Threat Research, 2025)
- 24% faster was developers’ pre-task forecast for AI-assisted completion. (Model Evaluation & Threat Research, 2025)
- 20% faster was developers’ post-study estimate of AI’s effect. (Model Evaluation & Threat Research, 2025)
- 19% slower was the observed completion-time effect when AI tools were allowed. (Model Evaluation & Threat Research, 2025)
- 39% faster was economists’ forecast before the observed result. (Model Evaluation & Threat Research, 2025)
- 38% faster was machine-learning experts’ forecast. (Model Evaluation & Threat Research, 2025)
- 143 hours of screen recordings were manually labeled. (Model Evaluation & Threat Research, 2025)
Enterprise adoption baseline statistics
- More than 97% had used AI coding tools at work at some point. (GitHub and Wakefield Research, 2024)
- 88% of U.S. respondents reported at least some company support for AI use. (GitHub and Wakefield Research, 2024)
- 59% of German respondents reported at least some company support. (GitHub and Wakefield Research, 2024)
- 30%–40% said their organization actively encouraged AI coding tools, depending on market. (GitHub and Wakefield Research, 2024)
- 29%–49% said tools were allowed but received limited encouragement, depending on market. (GitHub and Wakefield Research, 2024)
What the current evidence shows
- Developer surveys consistently show high adoption, yet trust in output accuracy has declined and “almost right” code remains a major source of friction.
- Self-reported productivity and observed productivity are not interchangeable: METR’s narrow randomized trial found experienced contributors were slower even though they believed AI had made them faster.
- The METR result should not be generalized to every developer or tool; it covers 16 highly experienced contributors using early-2025 systems in repositories they knew deeply.
- Tool percentages overlap because developers commonly use several assistants, editors, agents and general chatbots. They are adoption rates, not market shares.
- For current planning, verification, tests, security review and clear measurement remain necessary even when teams experience faster drafting or greater throughput.
In summary
| Statistic | Scope | Source | Year |
|---|---|---|---|
| 84% | used or planned to use AI tools in development, up from 76% in 2024. | Stack Overflow | 2025 |
| 51% | of professional developers used AI tools daily. | Stack Overflow | 2025 |
| 60% | expressed favorable sentiment toward AI tools, down from more than 70% in 2023 and 2024. | Stack Overflow | 2025 |
| 61% vs. 53% | of professionals versus learners expressed favorable sentiment. | Stack Overflow | 2025 |
| 46% | distrusted the accuracy of AI-tool output. | Stack Overflow | 2025 |
| 33% | trusted AI-tool accuracy. | Stack Overflow | 2025 |
| 3% | highly trusted AI-tool output. | Stack Overflow | 2025 |
| 20% | of experienced developers highly distrusted AI output. | Stack Overflow | 2025 |
| 64% | did not see AI as a threat to their jobs, down from 68% in 2024. | Stack Overflow | 2025 |
| 31% | currently used AI agents. | Stack Overflow | 2025 |
| 17% | planned to use AI agents. | Stack Overflow | 2025 |
| 38% | had no plans to use AI agents. | Stack Overflow | 2025 |
| 69% | of developers who used agents at work reported increased productivity. | Stack Overflow | 2025 |
| 52% | either did not use agents or stayed with simpler AI tools. | Stack Overflow | 2025 |
| 66% | were frustrated by AI solutions that were almost right but not quite. | Stack Overflow | 2025 |
| 45% | said debugging AI-generated code was more time-consuming. | Stack Overflow | 2025 |
| 75.3% | of respondents in the cited user subgroup did not trust AI answers. | Stack Overflow | 2025 |
| 61.7% | reported ethical or security concerns about code. | Stack Overflow | 2025 |
| 61.3% | wanted to understand their code fully. | Stack Overflow | 2025 |
| 35% | visited Stack Overflow because of AI-related issues at least some of the time. | Stack Overflow | 2025 |
| 36% | learned AI-enabled tools for work or career advancement. | Stack Overflow | 2025 |
| 44% | used AI tools to learn coding, up from 37%. | Stack Overflow | 2025 |
| 81% | of developers using LLMs reported OpenAI GPT models. | Stack Overflow | 2025 |
| 43% | reported Claude Sonnet models. | Stack Overflow | 2025 |
| 35% | reported Gemini Flash models. | Stack Overflow | 2025 |
| 18% | used Cursor. | Stack Overflow | 2025 |
| 10% | used Claude Code. | Stack Overflow | 2025 |
| 5% | used Windsurf. | Stack Overflow | 2025 |
| 90% | regularly used at least one AI tool at work for coding or development tasks. | JetBrains Research | 2026 |
| 74% | used a specialized AI developer tool rather than only a general chatbot. | JetBrains Research | 2026 |
| 76% | had heard of GitHub Copilot. | JetBrains Research | 2026 |
| 29% | used GitHub Copilot at work. | JetBrains Research | 2026 |
| 40% | of developers at companies with more than 5,000 employees used GitHub Copilot. | JetBrains Research | 2026 |
| 69% | were aware of Cursor. | JetBrains Research | 2026 |
| 18% | used Cursor at work. | JetBrains Research | 2026 |
| 57% | had heard of Claude Code in January 2026, up from 49% in September 2025 and 31% in mid-2025. | JetBrains Research | 2026 |
| 18% | used Claude Code at work worldwide. | JetBrains Research | 2026 |
| 24% | used Claude Code at work in the United States and Canada. | JetBrains Research | 2026 |
| 91% CSAT | was Claude Code’s reported satisfaction score. | JetBrains Research | 2026 |
| 54 NPS | was Claude Code’s reported likelihood-to-recommend score. | JetBrains Research | 2026 |
| 3% | used OpenAI Codex at work before the desktop app’s public launch. | JetBrains Research | 2026 |
| 6% | used Google Antigravity by January 2026. | JetBrains Research | 2026 |
| 28% | used the ChatGPT interface for coding tasks at work. | JetBrains Research | 2026 |
| 11% | used JetBrains AI Assistant and/or Junie. | JetBrains Research | 2026 |
| 16 developers | participated in the randomized controlled trial. | Model Evaluation & Threat Research | 2025 |
| 246 tasks | were completed in mature open-source projects. | Model Evaluation & Threat Research | 2025 |
| Five years | was participants’ average prior experience with their repository. | Model Evaluation & Threat Research | 2025 |
| 24% faster | was developers’ pre-task forecast for AI-assisted completion. | Model Evaluation & Threat Research | 2025 |
| 20% faster | was developers’ post-study estimate of AI’s effect. | Model Evaluation & Threat Research | 2025 |
| 19% slower | was the observed completion-time effect when AI tools were allowed. | Model Evaluation & Threat Research | 2025 |
| 39% faster | was economists’ forecast before the observed result. | Model Evaluation & Threat Research | 2025 |
| 38% faster | was machine-learning experts’ forecast. | Model Evaluation & Threat Research | 2025 |
| 143 hours | of screen recordings were manually labeled. | Model Evaluation & Threat Research | 2025 |
| More than 97% | had used AI coding tools at work at some point. | GitHub and Wakefield Research | 2024 |
| 88% | of U.S. respondents reported at least some company support for AI use. | GitHub and Wakefield Research | 2024 |
| 59% | of German respondents reported at least some company support. | GitHub and Wakefield Research | 2024 |
| 30%–40% | said their organization actively encouraged AI coding tools, depending on market. | GitHub and Wakefield Research | 2024 |
| 29%–49% | said tools were allowed but received limited encouragement, depending on market. | GitHub and Wakefield Research | 2024 |
Methodology note
This is a secondary synthesis of original-publisher research. Toolfountain did not conduct the underlying studies. Each statistic records its publisher, URL, publication date, evidence period, geography, sample population, methodology context, and verification date. Survey findings are not treated as causal measurements, and forecasts remain distinct from observed results.
Primary Sources
- 2025 Developer Survey — Stack Overflow (2025). More than 49,000 respondents
- Which AI Coding Tools Do Developers Actually Use at Work? — JetBrains Research (2026). More than 10,000 professional developers in the January 2026 wave
- Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity — Model Evaluation & Threat Research (2025). 16 developers completing 246 randomized tasks
- Survey: The AI Wave Continues to Grow on Software Development Teams — GitHub and Wakefield Research (2024). 2,000 non-manager enterprise developers and adjacent technical roles at companies with 1,000+ employees
