Vibe coding went from one tweet in February 2025 to a multi-billion-dollar category in about eighteen months, and the numbers around it now get quoted everywhere, usually without dates and often without sources. As of mid-2026, the vibe coding market is estimated at $4.7 billion to $7.06 billion depending on who’s measuring, 84% of developers use or plan to use AI coding tools, and roughly 41% of new code is AI-generated. Below are 30 vibe coding statistics with a source and a date attached to every single one. Where the sources disagree, and they do, we say so instead of quietly picking the flashier number.
Quick note on method: we only kept stats we could trace to a named study, survey, or company disclosure. Aggregator roundups are labeled as roundups, and nothing here is rounded differently than the original source printed it.
How big is the vibe coding market in 2026?
Nobody fully agrees on the size, because nobody fully agrees on what counts as vibe coding: just the prompt-to-app platforms, or every AI coding assistant on the planet. Hold that thought through every dollar figure in this section.
1. Vibe coding started as one tweet in February 2025
Andrej Karpathy, OpenAI co-founder and former Tesla AI director, posted on February 2, 2025 about a way of building software where you “fully give in to the vibes” and let the AI write the code. The post pulled more than 4.5 million views and accidentally named an industry, per Questera’s history of the term.
2. Nine months later, Collins Dictionary made it Word of the Year
Collins named “vibe coding” its 2025 Word of the Year, defining it as using natural-language prompts to have AI assist in writing computer code, as CNN reported in November 2025. Tweet to dictionary in nine months. That’s the adoption curve in one line.
3. The most-quoted market estimate says $4.7 billion for 2026, growing 38% a year
The figure you’ll see repeated most is $4.7 billion for 2026 with a 38% CAGR, heading for $12.3 billion by 2027, per Taskade’s 2026 State of Vibe Coding report. Credit where due: Taskade labels its own figures as estimates, which is more honesty than most market math gets.
4. A stricter count says $7.06 billion, growing 17.06% a year
Mordor Intelligence’s 2026 industry report sizes the market at $7.06 billion in 2026, on its way to $15.52 billion by 2031 at a 17.06% CAGR. Higher base, slower growth, different scope than Taskade’s number. Both get cited as gospel around the internet. They can’t both be right, so treat any single market figure as a range wearing a suit.
| Source | Market size | Growth rate | Projection |
|---|---|---|---|
| Taskade (industry estimate) | $4.7B in 2026 | 38% CAGR | $12.3B by 2027 |
| Mordor Intelligence | $7.06B in 2026 | 17.06% CAGR | $15.52B by 2031 |
| Business Research Company | $7.65B in 2025 | 23.8% CAGR | $22.2B by 2030 |
| Precedence Research | n/a (long-range) | 27.65% CAGR | $91.09B by 2035 |
5. Long-run forecasts land anywhere from $22.2 billion to $91.09 billion
The Business Research Company measured the adjacent AI coding market at $7.65 billion in 2025 and projects $22.2 billion by 2030 at a 23.8% CAGR, while Precedence Research goes further out: $91.09 billion by 2035 at 27.65%, both figures per Hostinger’s 2026 statistics index. When forecasts spread four times apart, the market is young and the crystal balls are foggy.
6. AI took 53% of all global venture funding in Q1 2025
Investors put $59.6 billion into AI in the first quarter of 2025, 53% of all global VC money, and the average vibe coding deal grew 71x between 2022 and 2025, from $7.4 million to $527.8 million, per the New Market Pitch data collected in the same Hostinger index. The money decided early.
Vibe coding statistics on adoption
Adoption is the one part of this story where every source points the same direction. The only argument is how far up.
7. 84% of developers use or plan to use AI coding tools
Stack Overflow’s 2025 Developer Survey put the number at 84%, up from 76% in 2024. That’s not early adopters anymore. That’s the profession.
8. By January 2026, 90% of developers used at least one AI tool at work
JetBrains’ AI Pulse survey found 90% of developers regularly using at least one AI tool as of January 2026, up from the 85% its State of Developer Ecosystem survey recorded in 2025, per Hostinger’s data index. The remaining 10% is a rounding error plus some very committed holdouts.
9. Yet 72% of professional developers say vibe coding is not part of their job
The same Stack Overflow 2025 survey found 72% of respondents saying vibe coding has no place in their professional workflow. Read items 7 and 9 together and the picture sharpens: developers use AI the way a chef uses a food processor. Constantly, and never unsupervised.
10. 63% of vibe coding users aren’t developers at all
Vercel’s data shows non-developers make up 63% of vibe coding users, as compiled in Hostinger’s 2026 index, and Hostinger’s own user research found 49% of them building business and portfolio websites. That’s exactly who our 40 free website prompts were written for: regular people building a real site, not engineers building the next Figma.
11. GitHub Copilot passed 20 million all-time users
Copilot crossed 20 million users as of July 2025, growing 75% year over year, and GitHub’s Octoverse data shows nearly 80% of new GitHub developers try Copilot within their first week, per the same 2026 roundup. The on-ramp to programming now has AI standing at the entrance.
12. A quarter of Y Combinator’s Winter 2025 batch shipped codebases that were 95% AI-generated
YC CEO Garry Tan said 25% of the Winter 2025 startups had products with 95% AI-generated code, and that the batch grew 10% per week in aggregate, per CNBC’s March 2025 report. These were technical founders who could have written it themselves. They chose not to.
How much code is AI-generated right now?
Company disclosures on this keep climbing, and they climb fast enough that the date on each number matters as much as the number itself.
13. Roughly 41% of all new code is AI-generated
The global aggregate estimate sits at 41% of new code, per Taskade’s 2026 roundup. Academic work backs the direction: a study published in Science found 29% of new Python functions on US GitHub were AI-generated by the end of 2024, via Hostinger’s index.
14. Gartner expects 60% of new code to be AI-generated by the end of 2026
Gartner projects the share hits 60% by the end of this year, per Taskade’s data. The same firm warns software defects could increase 2,500% by 2028 without proper governance, per Hostinger’s index. Volume up, guardrails optional. Hold that pairing in your head for the next section.
15. Google reports over 30% of its new code is AI-generated
Google’s figure passed 30% in Q1 2025, per the disclosure tracked in Hostinger’s index, up from the 25% recorded in Taskade’s roundup from Google’s earlier statements. Five points of growth in a few quarters, at Google scale, is an ocean of code.
16. Microsoft says 20% to 30% of code in some repositories is AI-written
Satya Nadella put the range at 20% to 30% for some Microsoft repositories in April 2025, in a CNBC interview cited in Hostinger’s data index. Notice the careful phrasing from the company that ships the world’s most-used coding assistant. The hedge is informative.
17. Anthropic says 70% to 90% of its code is AI-generated
At the far end of the disclosures, Anthropic told Fortune in January 2026 that 70% to 90% of its company code is AI-generated, per Hostinger’s roundup. The company building Claude mostly ships Claude-written code, which is either reassuring or unsettling depending on your week.
18. Copilot writes 46% of the code for developers who use it, and 61% for Java developers
GitHub’s own research found Copilot generating an average of 46% of code for its users, with Java developers reaching 61%, per Second Talent’s 2026 statistics roundup. Nearly half the code on a Copilot user’s screen was never typed by a human hand.
Vibe coding statistics on code quality, security, and trust
Here’s where the party gets quiet. These are the numbers the tool vendors leave out of the launch video.
19. AI models introduced security vulnerabilities in 45% of coding tasks
Veracode tested more than 100 large language models across 80 curated coding tasks and found AI-generated code introduced OWASP Top 10 security vulnerabilities in 45% of cases, per its July 2025 GenAI Code Security Report. Java was the worst offender, failing security tests more than 70% of the time.
20. Security pass rates haven’t improved since 2023
Veracode’s follow-up research found syntax pass rates climbed from about 50% to 95% since 2023 while security pass rates stayed flat between 45% and 55% regardless of model generation, per its Spring 2026 update. The models got fluent much faster than they got safe. That one sentence explains most of the cleanup work in this industry.
21. AI-generated code carries 1.7x more issues than human-written code
CodeRabbit’s review of 470 pull requests found AI-generated code with 1.7x more major issues and cross-site scripting vulnerabilities running 2.74x higher, per the study cited in Taskade’s 2026 report. Not catastrophic. Not free, either.
22. 10.3% of audited Lovable apps shipped with a critical security flaw
Researchers found 170 of 1,645 checked Lovable-built apps, 10.3%, exposed through broken row-level security, logged as CVE-2025-48757, per Taskade’s roundup. One in ten apps looked finished while the database sat open. It’s the same pattern we keep meeting in rescue work: the front end is gorgeous, the back door is missing.
23. Developer trust in AI accuracy fell to 29%
Trust in the accuracy of AI output dropped from roughly 40% to 29% year over year, with 46% of developers actively distrusting it against 33% who trust it, and only 3% reporting high trust, per Stack Overflow’s 2025 survey press release. The top frustration, named by 45% of respondents: AI answers that are “almost right, but not quite.”
24. 96% of developers don’t fully trust AI code, but only 48% always review it
SonarSource found 96% of developers don’t fully trust AI-generated code to be functionally correct, while just 48% always review it before committing, per the 2025 survey tracked in Hostinger’s index. Everybody’s shipping code they don’t quite believe in. That gap is where the cleanup industry lives.
Productivity and jobs: the numbers that fight each other
Careful with this section. It’s where cherry-picking runs wildest, so you’re getting both sides with the methodology attached.
25. Experienced developers were 19% slower with AI tools, and thought they were 20% faster
METR’s randomized controlled trial had 16 experienced open-source developers complete 246 real tasks, and AI tools made them 19% slower. They predicted a 24% speedup going in, and still believed they’d been 20% faster after it was over, as The Register reported in July 2025. A 39-point gap between feeling and fact.
26. Other trials found 21% to 26% gains, plus 41% more bugs
An MIT randomized trial across 4,867 developers measured a 26% increase in completed tasks, Google’s internal study clocked 21% faster completion, and a controlled JavaScript task ran 55.8% faster, while Uplevel’s observational study of roughly 800 developers found 41% more bugs, all per the studies gathered in Hostinger’s index. The honest synthesis: AI speeds up simple, fresh work and slows down experts inside old, tangled codebases. Both findings are true at once.
27. Entry-level developer employment fell nearly 20% from its late-2022 peak
Employment for entry-level developers declined nearly 20% between late 2022 and July 2025, per Stack Overflow data compiled in Hostinger’s index. The junior rungs of the ladder are where AI bites first, which is worth remembering whenever someone insists nothing is changing.
The money behind the vibe coding tools
Follow the revenue and the ambiguity thins out fast. Whatever the productivity studies argue about, buyers already voted.
28. Cursor went from $1.2 billion to $4 billion in annualized revenue in five months
Cursor ended 2025 at $1.2 billion ARR and hit $4.0 billion by May 2026, after a November 2025 Series D of $2.3 billion at a $29.3 billion valuation, per Sacra’s company profile, which calls it the fastest-growing SaaS company of all time. In June 2026, SpaceX signed an agreement to acquire it for $60 billion.
29. Lovable hit $100 million ARR in eight months, then doubled four months later
Lovable reached $100 million in ARR within eight months of launch and $200 million four months after that, raising $330 million at a $6.6 billion valuation in December 2025 with 25 million projects created, per the TechCrunch reporting collected in Hostinger’s index. Eight months. This industry used to call that a beta period.
30. Replit booked $240 million in 2025 and is targeting $1 billion ARR by the end of 2026
Replit generated $240 million in 2025 revenue, raised $400 million at a $9 billion valuation in March 2026, and set a $1 billion ARR target for year-end, while OpenAI’s Codex counted 4 million weekly active developers in April 2026, per Hostinger’s 2026 index. The tools cashing these checks are the same ones failing 45% of Veracode’s security tests. Both facts fit inside one market.
What the stats mean if you’re deciding between DIY and hiring
Strip the hype off those 30 numbers and three things stay standing. The tools work well enough that most users aren’t developers anymore. The security and quality risks are real, measurable, and flat since 2023. And the gap between “looks done” and “is done” is exactly where small businesses get hurt, because item 24 applies double when nobody on the team reads code at all.
For a simple marketing site, the math favors DIY more than the scary stats suggest. The 45% vulnerability figure comes from application code: logins, databases, payment flows. A five-page site with a contact form sits in the shallow end of that risk pool, and with a careful prompt and some patience you can absolutely build it yourself. We publish 40 free prompts precisely because that’s true.
Once real money runs through the thing, the calculus flips. If you want the full cost math, DIY tools versus flat-fee builds and where each one wins, we broke it down in what a vibe-coded website actually costs. And if you’re wondering whether an AI-built site can even rank, the short answer is that Google judges the page, not the author. The long answer, with receipts, is in do AI websites rank on Google.
One more thing worth noticing: this post is a stats page, and stats pages are what AI engines quote. That’s not an accident, it’s the playbook we documented in how to get cited by AI search. If you’d rather have your business show up inside ChatGPT and Google’s AI answers than just read about other people’s numbers, that’s the exact job of our SEO and AI-search service: $1,200 for the audit, ongoing work from $950 a month, and if it’s not a fit for your situation, we’ll tell you within a day.
Questions we get about these numbers
How big is the vibe coding market in 2026?
Somewhere between $4.7 billion and $7.06 billion, depending on whose scope you accept. Taskade’s estimate puts it at $4.7 billion growing at a 38% CAGR toward $12.3 billion by 2027, while Mordor Intelligence measures $7.06 billion growing 17.06% annually toward $15.52 billion by 2031. The spread comes from definitions: narrow counts include only prompt-to-app platforms, broad counts fold in the whole AI-assisted development category. Treat any single figure as an estimate with error bars.
What percentage of code is AI-generated?
Aggregate estimates put it around 41% of new code globally in 2026. Company disclosures range wider: Google reports over 30%, Microsoft says 20% to 30% in some repositories, GitHub Copilot users see 46% of their code AI-written on average, and Anthropic reports 70% to 90%. Gartner projects 60% of all new code will be AI-generated by the end of 2026, so every one of these numbers has a short shelf life.
Who coined the term vibe coding?
Andrej Karpathy, OpenAI co-founder and former Tesla AI director, in a post on February 2, 2025 describing a style of building where you “fully give in to the vibes” and let AI write the code. The post drew more than 4.5 million views, and by November 2025 Collins Dictionary had named vibe coding its Word of the Year.
Is AI-generated code safe to use without review?
No. Veracode’s testing of more than 100 models found security vulnerabilities introduced in 45% of coding tasks, and its Spring 2026 follow-up showed security pass rates flat since 2023 even as code fluency improved. The risk scales with what the code touches: a static brochure page is low-stakes, while anything handling logins, payments, or customer data needs a human who can actually read what the AI wrote.
Do these statistics mean I shouldn’t build my own website with AI?
Not at all. Most of the alarming numbers measure application code at company scale, not five-page small-business sites. For a simple site, vibe coding with good prompts works, and the free route is a legitimate route. The stats argue for knowing your handoff point: once your site takes payments, stores customer data, or becomes the thing your revenue depends on, the cost of a professional build gets cheap compared to the cost of stat number 22 happening to you.