Research · Developer survey
State of Vibecoding 2026: Developer Survey
AI writes more code than ever. We want to measure how often developers and students ship code they can't explain, how they check it, and what it's doing to their fundamentals.
It takes about 4 minutes and it's anonymous. Results will be published openly under CC BY 4.0, and every respondent can get the report by email (optional).
- About 4 minutes
- Anonymous
- 14 questions
- Results under CC BY 4.0
About this survey
The State of Vibecoding 2026 survey asks developers, students and AI builders 14 short questions about AI-generated code: how much of their code AI writes, how often they commit code they couldn't explain line by line, and how they verify it. The survey is open now; no results have been published yet. More Contral research.
Take the survey
Answer honestly; there are no right answers. Your progress isn't saved, so finish in one go.
What we'll publish
An open report, not a marketing deck
Once enough people have answered, we'll publish the results here. Until then there are no numbers to share.
Charts and tables for every question
Broken down by role and years of experience wherever there are enough responses to report a group on its own.
The full questionnaire
Exact question wording and answer options, as listed on this page, so you can judge the results for yourself.
Methodology and limits
How many people answered, how we found them, how we cleaned the data, and what a self-selected online survey can and can't tell you.
Free to cite and reuse
Results are released under CC BY 4.0. Use them in articles, talks, papers and courses with attribution and a link to this page.
Get notified
Get the report by email
Add your email on the last step of the survey and we'll send you the report when it's published. That's the only thing we'll use it for. Journalists and researchers who want to hear when results are out can email contact@contral.ai.
Every response makes the results more useful. Sharing the survey with your team, class or community helps more than anything else.
Share the survey
The questionnaire
Every question we ask
Published in full so the results can be read in context.
Which best describes you?
Student (school or university) · Bootcamp student or graduate · Self-taught, learning to code · Junior developer (under 2 years professionally) · Mid-level developer · Senior or staff developer · Not a developer, but I build things with AI
How long have you been writing code, in any form?
Less than 1 year · 1–2 years · 3–5 years · 6–10 years · More than 10 years
Where are you based?(optional)
Country list, or prefer not to say
Which AI coding tools have you used in the last 3 months?
GitHub Copilot · Cursor · Claude Code · ChatGPT · Windsurf · Gemini · Replit · Contral · Other · None
Roughly how much of the code you wrote last month was generated by AI?
0–10% · 10–25% · 25–50% · 50–75% · 75–100%
How often do you commit or submit AI-generated code you couldn't explain line by line?
Never · Rarely · Sometimes · Often · Most of the time · Not applicable: I don't commit or submit code
Think of the last change an AI tool made to your code. How confident are you that you could explain it line by line?
1 (Not at all confident) to 5 (Completely confident)
Have you ever shipped or submitted a bug that came from AI-generated code you didn't fully understand?
Yes · No · Not sure
Has anyone (in an interview, code review, class or exam) asked you to explain code you couldn't explain?
Yes · No · It hasn't come up: no interviews, reviews or exams yet
Overall, has AI made you better or worse at programming fundamentals?
1 (Much worse) to 5 (Much better)
How do you usually check AI-generated code before you keep it?
Read it line by line · Write or run tests · Run it and see if it works · Ask the AI to explain it · Usually nothing
Would you use a tool that quizzes you on the AI-generated code in your own projects?
1 (Definitely not) to 5 (Definitely)
What's your biggest worry about relying on AI to write code?(optional)
Free text, up to 500 characters
Email for the report(optional)
With a consent checkbox. Stored separately from your answers.
Methodology and privacy
How we handle your answers
- Anonymous. No sign-in, and we don't ask for your name. We store your answers with a one-way hash of your IP address, used only to limit duplicate submissions. We don't store the IP address itself.
- Aggregated only. We publish totals and percentages, never individual responses. If we quote a free-text answer, we quote it without anything that could identify you.
- Email is optional. If you leave one, it goes into a separate list with no ID linking it to your answers, and we use it only to send you the report.
- Who's asking. The survey is run by Contral, which makes an AI coding assistant that teaches. We'll disclose that alongside the results, and we'll report the limits of a self-selected online sample.
- Read the full privacy policy.
FAQ
About the survey
- Who is running the State of Vibecoding 2026 survey?
- Contral, which makes an AI coding assistant that teaches. Because Contral builds tools that quiz people on AI-generated code, one question asks about interest in that kind of tool; we'll report it with that disclosure.
- Is the survey anonymous?
- Yes. There's no sign-in and we don't ask for your name. Answers are stored with a one-way hash of your IP address, used only to limit duplicate submissions. The optional email for the report is stored in a separate list with no ID linking it to your answers.
- Who should take it?
- Anyone who writes code: students, bootcamp learners, self-taught programmers, professional developers at any level, and people who aren't developers but build things with AI. You don't need to use AI tools to answer.
- When will the results be published?
- When we have enough responses for meaningful breakdowns. Leave your email at the end of the survey and we'll send you the report when it's out.
- Can I use the results in my article, paper or course?
- Yes. Results will be published under the Creative Commons Attribution 4.0 (CC BY 4.0) license: reuse them freely with attribution to Contral and a link to this page.