Live on AppSumo: from $79 (opens AppSumo in a new tab)

Research · Data roundup

AI Coding Statistics 2026

68 verified statistics on how developers use AI to write code: adoption, productivity, code quality and security, trust, learning, junior developer jobs, vibecoding, CS enrollment and the market. Every number links to its original source.

Last updated: · 42 sources · Cite this page

Headline stats

AI coding tool adoption

Every major 2025–2026 developer survey now puts AI use among developers somewhere between 80% and 90%. In 2026 the question shifted from whether developers use AI to how much of the work they hand to agents.

90%

#

of professional developers use AI coding agents at work at least weekly

JetBrains' Developer Ecosystem Survey 2026 (15,000+ professional developers, fielded May–July 2026) found 90% using local or cloud coding agents at least weekly and 68% using them daily.

Source: JetBrains Research, 2026

73%

#

of AI coding assistant/agent users use them daily

In Stack Overflow's 2026 Developer Survey, 73% of respondents who use AI coding assistants or agents said they use them every day.

Source: Stack Overflow, 2026

31%

#

of daily AI users spend 4+ hours a day using AI tools

Stack Overflow's 2026 survey found 31% of daily AI users use AI tools 4+ hours per day, 24% for 2–4 hours and 26% for 1–2 hours.

Source: Stack Overflow, 2026

84%

#

of developers use or plan to use AI tools

84% of Stack Overflow 2025 Developer Survey respondents were using or planning to use AI tools in their development process, up from 76% in 2024.

Source: Stack Overflow, 2025

51%

#

of professional developers use AI tools daily

Stack Overflow's 2025 survey reported that 51% of professional developers use AI tools every day.

Source: Stack Overflow, 2025

62%

#

of developers were already using AI tools in 2024

Stack Overflow's 2024 survey found 62% of respondents currently using AI tools, up from 44% the year before, with 76% using or planning to use them.

Source: Stack Overflow, 2024

85%

#

of developers regularly use AI for coding and development

JetBrains' State of Developer Ecosystem 2025 (24,534 developers, 194 countries) found 85% regularly use AI tools and 62% rely on at least one AI coding assistant, agent, or code editor.

Source: JetBrains, 2025

90%

#

AI adoption among software development professionals

Google's 2025 DORA report (nearly 5,000 technology professionals) put AI adoption at 90%, a 14% increase from 2024, with a median of two hours a day spent working with AI.

Source: Google Cloud / DORA, 2025

40%+

#

of GitHub pull requests started with an agent's first draft

GitHub's Octoverse 2026 reports that in September 2026 more than 40% of pull requests on GitHub started with an agent's first draft, up from 17% in March 2026.

Source: GitHub, 2026

55%

#

of GitHub Copilot monthly active users now use agentic features

Per Octoverse 2026, 55% of GitHub Copilot's monthly active users use agentic features, up from 35% a year earlier.

Source: GitHub, 2026

80%

#

of new GitHub developers use Copilot in their first week

GitHub's Octoverse 2025 reported that 80% of new developers on GitHub use Copilot in their first week, as 36 million+ developers joined in a year.

Source: GitHub, 2025

39%

#

of professional developers use Claude Code at work

JetBrains' 2026 survey found Claude Code use at work rose from 18% (January 2026) to around 39% (May–July 2026), while GitHub Copilot fell from 29% a year earlier to 21%.

Source: JetBrains Research, 2026

90%

#

of enterprise software engineers will use AI code assistants by 2028 (forecast)

Gartner predicts that by 2028, 90% of enterprise software engineers will use AI code assistants, up from less than 14% in early 2024. This is a forecast, not a measurement.

Source: Gartner (via GitHub Blog), 2025

How much code AI writes

Company and survey figures on the share of code written by AI are self-reported and measured differently, so treat them as directional, not comparable.

~47%

#

of developers' work code is fully written by AI agents (self-reported average)

Professional developers in JetBrains' 2026 survey reported on average that ~47% of their code was fully agent-written, ~38% written with some AI assistance and ~27% fully manual (bucketed answers, so the averages can exceed 100%).

Source: JetBrains Research, 2026

~22%

#

of developers get over 80% of their code from agents

In JetBrains' 2026 data, developers relying almost entirely on agents (over 80% agent-generated code) remain a minority of around 22%, while one in five writes zero code without AI help.

Source: JetBrains Research, 2026

25%+

#

of all new code at Google was generated by AI (Oct 2024)

Sundar Pichai said on Alphabet's Q3 2024 earnings call that more than a quarter of all new code at Google is generated by AI, then reviewed and accepted by engineers.

Source: Google (Alphabet), 2024

20–30%

#

of the code in Microsoft's repos is written by AI, per its CEO

At Meta's LlamaCon on April 29, 2025, Satya Nadella said "maybe 20%, 30% of the code that is inside of our repos today" is written by AI.

Source: CNBC, 2025

32% vs 13%

#

of senior vs junior developers say over half their shipped code is AI-generated

Fastly's July 2025 survey of 791 professional developers found 32% of seniors (10+ years) and 13% of juniors (0–2 years) say over half of their shipped code is AI-generated.

Source: Fastly, 2025

79%

#

of Claude Code conversations were classified as automation

Anthropic's Economic Index (April 2025) classified 79% of Claude Code conversations as "automation" (AI performing the task directly), versus 49% of Claude.ai conversations.

Source: Anthropic, 2025

Productivity: what controlled studies show

Controlled experiments disagree, and the disagreement is the finding: results depend heavily on task type, developer experience, codebase familiarity and the tool generation tested.

19%

#

slower: experienced open-source developers took longer with early-2025 AI tools

METR's July 2025 randomized trial (16 experienced open-source developers, 246 real issues in their own repos) found that when AI tools were allowed, developers took 19% longer to finish.

Source: METR, 2025

+20%

#

speedup developers believed they got, despite measurably slowing down

In METR's 2025 study, developers expected AI to speed them up by 24% and, even after the measured 19% slowdown, still believed AI had sped them up by 20%.

Source: METR, 2025

−18%

#

METR's late-2025 point estimate for returning developers — with an interval that still includes a slowdown

METR's February 2026 update (57 developers, 800+ tasks) estimated a "speedup of -18%" (95% CI −38% to +9%) for returning developers and −4% (CI −15% to +9%) for new recruits, calling the data "only very weak evidence" because of selection effects.

Source: METR, 2026

55.8%

#

faster task completion with GitHub Copilot in a lab experiment

A 2023 controlled experiment found developers with GitHub Copilot implemented an HTTP server in JavaScript 55.8% faster than the control group — a single, well-defined greenfield task.

Source: arXiv (Peng et al.), 2023

26.08%

#

more completed tasks across three corporate field experiments

Pooled RCTs at Microsoft, Accenture and a Fortune 100 company (4,867 developers) found a 26.08% increase (SE: 10.3%) in completed tasks with an AI coding assistant; less experienced developers gained more.

Source: Cui, Demirer, Jaffe, Musolff, Peng & Salz (Management Science), 2025

~21%

#

faster on an enterprise task in Google's internal RCT

A randomized controlled trial with 96 full-time Google software engineers estimated the time reduction from AI features at about 21%, though the authors note the confidence interval is large.

Source: arXiv (Google), 2024

80%+

#

of tech professionals say AI has enhanced their productivity

In Google's 2025 DORA research, over 80% of respondents said AI has enhanced their productivity and 59% reported a positive influence on code quality (self-reported).

Source: Google Cloud / DORA, 2025

1 in 5

#

developers using AI save 8+ hours a week

JetBrains' 2025 survey found nearly nine out of ten developers using AI save at least an hour every week, and one in five saves eight hours or more (self-reported).

Source: JetBrains, 2025

28%

#

of developers say fixing AI code offsets most of their time savings

In Fastly's 2025 survey, 28% of developers said they often have to fix or edit AI-generated code enough to offset most of the time savings.

Source: Fastly, 2025

Code quality and security of AI-generated code

Benchmarks and audits keep finding that AI-generated code needs real review — especially for security — and that AI users can be overconfident in it.

45%

#

of AI coding test cases introduced a security vulnerability

Veracode's 2025 GenAI Code Security Report tested 80 curated coding tasks across more than 100 LLMs and found AI introduced security vulnerabilities in 45 percent of cases.

Source: Veracode, 2025

70%+

#

security failure rate for AI-generated Java

Java was the riskiest language in Veracode's 2025 testing, with a security failure rate over 70 percent.

Source: Veracode, 2025

86%

#

of the time, LLMs failed to defend against cross-site scripting

In Veracode's tests, LLMs failed to secure code against cross-site scripting (CWE-80) in 86% of cases and against log injection (CWE-117) in 88%.

Source: Veracode, 2025

~1.7×

#

more issues in AI-co-authored pull requests

CodeRabbit's analysis of 470 open-source GitHub PRs (December 2025) found AI-co-authored PRs averaged 10.83 issues versus 6.45 for human-only PRs; authorship was inferred, so labels are not guaranteed.

Source: CodeRabbit, 2025

2.74×

#

higher rate of security issues in AI-co-authored PRs (upper bound)

CodeRabbit reported security issues were up to 2.74× higher and readability issues more than 3× higher in AI contributions.

Source: CodeRabbit, 2025

8×

#

increase in duplicated code blocks during 2024

GitClear's 2025 research (211 million changed lines) found code blocks with 5+ duplicated lines increased 8× during 2024, and 2024 was the first year copy/pasted lines exceeded moved (refactored) lines.

Source: GitClear (via DevClass), 2025

~40%

#

of Copilot-generated programs were vulnerable in security-relevant scenarios

NYU researchers prompted GitHub Copilot with 89 security-relevant scenarios, producing 1,689 programs, and found approximately 40% to be vulnerable (2021 study, early Copilot).

Source: NYU (Pearce et al., IEEE S&P 2022), 2021

47

#

participants: AI users wrote less secure code yet felt more secure

In a Stanford user study with 47 participants, those with an AI assistant wrote significantly less secure code and were more likely to believe they wrote secure code than those without one.

Source: Stanford (Perry et al., ACM CCS 2023), 2023

Trust and developer sentiment

Use keeps rising while trust lags. Developers increasingly use AI daily without trusting its output by default.

46%

#

of developers distrust the accuracy of AI tools

In Stack Overflow's 2025 survey, 46% of respondents said they distrust the accuracy of AI tools versus 33% who trust it; only about 3% "highly trust" AI output.

Source: Stack Overflow, 2025

66%

#

cite "AI solutions that are almost right, but not quite" as a top frustration

The most common frustration in Stack Overflow's 2025 survey was AI answers that are almost right but not quite (66%); 45% said debugging AI-generated code is more time-consuming.

Source: Stack Overflow, 2025

60%

#

favorability toward AI tools in 2025, down from 70%+

Positive sentiment toward AI tools fell to about 60% in Stack Overflow's 2025 survey, from over 70% in 2023 and 2024.

Source: Stack Overflow, 2025

62%

#

of developers view AI at least somewhat positively in 2026

Stack Overflow's 2026 survey found 62% of all respondents view AI at least somewhat positively, rising to 71% among daily users.

Source: Stack Overflow, 2026

48%

#

trust AI when they can easily validate the answer

Asked when trusting AI is acceptable, 48% of Stack Overflow 2026 respondents said when they can easily validate the answers.

Source: Stack Overflow, 2026

17%

#

avoid AI at work or school over concern about losing job skills

In Stack Overflow's 2026 survey, 6 in 10 respondents cited a reason to avoid using AI at work or school; the top one was concern over losing job skills or training an AI to replace you (17%).

Source: Stack Overflow, 2026

30%

#

of tech professionals trust AI-generated output only a little or not at all

Google's 2025 DORA research found 30% trust AI "a little" (23%) or "not at all" (7%).

Source: Google Cloud / DORA, 2025

Learning and skill formation

The strongest evidence on learning says the same thing across settings: how you use AI matters more than whether you use it. Delegating the thinking boosts practice scores but hurts later mastery.

17 pts

#

lower comprehension scores for developers who learned a new library with AI

Anthropic's January 2026 randomized trial (52 mostly junior engineers) found the AI group averaged 50% on a comprehension quiz versus 67% for hand-coders (p = 0.01); the AI group's ~2-minute speed advantage was not statistically significant.

Source: Anthropic, 2026

65%+ vs <40%

#

quiz scores: conceptual AI use vs delegating to AI

In the same Anthropic study, high-scoring interaction patterns (asking conceptual questions, requesting explanations, or generating code and then working to understand it) averaged 65% or higher, while patterns that leaned on AI to write or debug the code averaged under 40%. These are associations, not causal estimates.

Source: Anthropic, 2026

−17%

#

exam performance after unrestricted AI practice (vs. no AI)

In a field experiment with nearly 1,000 Turkish high school math students, an unrestricted GPT-4 tutor raised practice performance 48% but lowered exam performance 17% versus control; a guarded tutor avoided the drop. (Math, not coding.)

Source: Knowledge at Wharton (Bastani et al., PNAS 2025), 2024

44%

#

learned to code with AI tools in the past year

44% of Stack Overflow 2025 respondents said they used AI CodeGen tools or AI-enabled apps to learn to code in the past year.

Source: Stack Overflow, 2025

53%

#

of people learning to code view AI tools favorably

Only 53% of respondents learning to code were favorable toward AI tools in Stack Overflow's 2025 survey, versus 61% of professional developers.

Source: Stack Overflow, 2025

36.8%

#

of students' Claude conversations were Computer Science

Anthropic's April 2025 Education Report found Computer Science accounted for 36.8% of students' conversations while comprising only 5.4% of U.S. degrees; ~47% of student conversations were "Direct" (seeking answers with minimal engagement).

Source: Anthropic, 2025

50%

#

of developers are ready to delegate writing application logic to AI

JetBrains' 2026 data shows readiness to delegate application logic rose from 18% (April–June 2025) to 50% (May–July 2026), while 54% are still not ready to delegate debugging.

Source: JetBrains Research, 2026

319

#

knowledge workers: more trust in GenAI, less critical thinking

A Microsoft Research survey of 319 knowledge workers (936 real examples, CHI 2025) found higher confidence in GenAI is associated with less critical thinking, while higher self-confidence is associated with more.

Source: Microsoft Research (ACM CHI 2025), 2025

Junior developers and jobs

Overall developer employment has not collapsed, but entry-level hiring has been hit hardest. Causality is still debated; these are the most-cited measurements.

19%

#

employment gap for 22–25-year-olds in AI-exposed jobs

Using ADP payroll data through June 2026, Stanford researchers found employment of workers aged 22–25 in AI-exposed occupations stands 19% below where it would be had it kept pace with less-exposed peers; experienced workers show no comparable gap.

Source: Stanford Digital Economy Lab, 2026

7%

#

of Big Tech hires are new grads

SignalFire's 2025 State of Tech Talent report found new grads account for just 7% of Big Tech hires, with new hires down 25% from 2023 and over 50% from pre-pandemic 2019 levels.

Source: SignalFire, 2025

80.8

#

Indeed U.S. software development postings index (Feb 2020 = 100)

Indeed's U.S. software development job postings index stood at 80.79 on October 23, 2026 — about 19% below its February 1, 2020 baseline.

Source: Indeed Hiring Lab (via FRED, St. Louis Fed), 2026

37%

#

of developers think AI will decrease demand for developers

In JetBrains' 2026 research, 37% of developers believe AI coding adoption will decrease demand for professional developers, up from 28% in 2024.

Source: JetBrains Research, 2026

79%

#

expect employers to require AI-tool proficiency

79% of developers in JetBrains' 2026 research believe employers will expect proficiency with AI dev tools, up from 59% in June–July 2024.

Source: JetBrains Research, 2026

70%

#

of professional developers did not see AI as a threat to their job (2024)

In Stack Overflow's 2024 survey, 70% of professional developers said they do not perceive AI as a threat to their job.

Source: Stack Overflow, 2024

Vibecoding

"Vibe coding" went from a February 2025 post to a dictionary's Word of the Year in nine months — but most professional developers say it is not how they work.

2025

#

Collins Word of the Year: "vibe coding"

Collins Dictionary named vibe coding its Word of the Year 2025, defining it as the use of artificial intelligence prompted by natural language to write computer code.

Source: Collins Dictionary, 2025

72%

#

of developers say vibe coding is not part of their professional work

In Stack Overflow's 2025 survey, 72% said vibe coding (generating software from LLM prompts) is not part of their professional development work, and an additional 5% emphatically said no.

Source: Stack Overflow, 2025

25%

#

of YC's W25 startups had 95% AI-generated codebases

Y Combinator managing partner Jared Friedman said a quarter of the Winter 2025 batch have 95% of their codebases generated by AI (founder self-reports).

Source: TechCrunch, 2025

Computer science education and enrollment

After a decade-long boom, computing enrollment is cooling — even as degree production hits records from earlier cohorts.

−13%

#

new CS bachelor's majors (2025 Taulbee Survey)

CRA's 2025 Taulbee Survey found new CS majors dropped to 33,490, down 13 percent across its longitudinal cohort of departments; total bachelor's enrollment fell 4%.

Source: Computing Research Association (CRA), 2026

41,858

#

CS bachelor's degrees awarded — a record

The same CRA Taulbee Survey reported CS bachelor's degree production rose to a record 41,858 in 2025, up 13 percent for the longitudinal cohort, as earlier boom cohorts graduated.

Source: Computing Research Association (CRA), 2026

62%

#

of computing departments reported falling undergraduate enrollment

In CRA's fall 2025 pulse survey of 134 academic units, 62% reported decreased bachelor's computing enrollment for 2025–26, 19% steady and 13% increased; leaders cited job-market and AI fears among the reasons.

Source: Computing Research Association (CRA), 2025

−8.4%

#

undergraduate computer science enrollment at four-year institutions (spring 2026)

National Student Clearinghouse data show undergraduate Computer and Information Sciences enrollment fell 8.4% at four-year institutions in spring 2026, while overall undergraduate enrollment rose 1.3%.

Source: National Student Clearinghouse, 2026

Market size and business

Analyst "market size" estimates vary wildly, so we list only revenue and user figures disclosed by the companies themselves.

$1B

#

Claude Code run-rate revenue, six months after public launch

Anthropic said in December 2025 that Claude Code reached $1 billion in run-rate revenue just six months after becoming available to the public.

Source: Anthropic, 2025

$1B+

#

Cursor annualized revenue (Nov 2025)

Cursor's maker Anysphere said in November 2025 it had crossed $1B in annualized revenue, announcing a $2.3B Series D at a $29.3B post-money valuation.

Source: Cursor (Anysphere), 2025

20M+

#

all-time GitHub Copilot users (July 2025)

Microsoft said GitHub Copilot passed 20 million all-time users in July 2025 and is used by 90% of the Fortune 100.

Source: TechCrunch, 2025

1.1M+

#

public GitHub repositories use an LLM SDK

GitHub's Octoverse 2025 counted more than 1.1 million public repositories using an LLM SDK.

Source: GitHub, 2025

What this means for developers learning to code

AI is now the default way code gets written. With 90% of professional developers using coding agents weekly and agents drafting more than 40% of GitHub pull requests, learning to code without AI is no longer realistic preparation for the job. The question for learners is not whether to use AI, but how.

Speed and understanding are different outcomes. The productivity evidence is genuinely mixed, but the learning evidence is consistent: when AI does the thinking, people do better in the moment and worse afterwards. Anthropic's trial found a 17-point comprehension gap, and a high-school math experiment saw practice scores rise 48% while exam scores fell 17%. In both studies, the damage shrank when AI was used to explain rather than to hand over answers: conceptual questions scored 65%+ in Anthropic's trial, and a tutor with guardrails largely avoided the exam drop.

Review skill is the new baseline. Benchmarks find security flaws in a large share of AI-generated code, and users with AI help are often more confident in insecure code. Entry-level hiring is also the part of the market under the most pressure, so juniors increasingly need to show they can read, debug and defend code they did not type. Our guide to understanding an AI-generated codebase covers that skill in practice.

Where Contral fits. We build Contral around this evidence. Build Mode lets you work with an agent at full speed, Learn Mode teaches the concepts behind the code, and Defense Mode asks you to explain what was just built before you ship it, the kind of active engagement associated with the higher comprehension scores above. For more context, read our breakdown of the Anthropic skills-gap study, what vibecoding gets right and wrong, and the terms in our AI coding glossary.

Methodology and sources

How we chose stats. We prioritized large, repeated industry surveys (Stack Overflow, JetBrains, GitHub Octoverse, Google DORA), randomized controlled trials, peer-reviewed research and company disclosures, published 2024–2026. Older studies are included only where they remain the most-cited evidence on a question, and are dated as such.

How we verified them. Each number was checked against its primary source and is reported with the source's own wording, sample and date. We do not round numbers or convert percentage-point gaps into relative percentages. Where a primary page blocks automated access, we verified the figure through a named, reputable outlet that reports it and flagged that below. Figures we could not verify were left out.

How to read them. Survey figures are self-reported and samples differ (for example, Stack Overflow's respondents skew toward its own community). Company figures on the "share of code written by AI" are not measured the same way and should not be compared directly. Vendor-published research (Veracode, CodeRabbit, GitClear, Fastly) is labeled with the vendor's name.

Corrections. Spotted an error or a newer dataset? Contact us and we will review it in the next update.

All sources (42)

  1. Anthropic (2026). How AI assistance impacts the formation of coding skills
  2. Anthropic (2025). Anthropic Economic Index: AI's impact on software development
  3. Anthropic (2025). Anthropic Education Report: How University Students Use Claude
  4. Anthropic (2025). Anthropic acquires Bun as Claude Code reaches $1B milestone
  5. arXiv (Google) (2024). How much does AI impact development speed? An enterprise-based randomized controlled trial
  6. arXiv (Peng et al.) (2023). The Impact of AI on Developer Productivity: Evidence from GitHub Copilot
  7. CNBC (2025). Satya Nadella says as much as 30% of Microsoft code is written by AI
  8. CodeRabbit (2025). State of AI vs. Human Code Generation Report
  9. Collins Dictionary (2025). Collins' Word of the Year 2025: AI meets authenticity as society shifts
  10. Computing Research Association (CRA) (2026). New CRA Taulbee Survey Findings Show Record Degree Production Alongside a Cooling Enrollment Pipeline
  11. Computing Research Association (CRA) (2025). CERP Pulse Survey: A Snapshot of 2025 Undergraduate Computing Enrollment Patterns
  12. Cui, Demirer, Jaffe, Musolff, Peng & Salz (Management Science) (2025). The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers
  13. Cursor (Anysphere) (2025). Series D announcement
  14. Fastly (2025). Vibe Shift? Senior Developers Ship nearly 2.5x more AI Code than Junior Counterparts
  15. Gartner (via GitHub Blog) (2025). Gartner positions GitHub as a Leader in the 2025 Magic Quadrant for AI Code AssistantsGartner prediction quoted by GitHub; the underlying Magic Quadrant report is paywalled.
  16. GitClear (via DevClass) (2025). AI is eroding code quality states new in-depth reportGitClear's own report page blocks automated access; figures verified via DevClass's coverage of the report.
  17. GitHub (2026). Octoverse 2026: Commits, pushes reach new highs as developers build with agents
  18. GitHub (2025). Octoverse 2025: A new developer joins GitHub every second as AI leads TypeScript to #1
  19. Google (Alphabet) (2024). Q3 2024 earnings call: CEO Sundar Pichai's remarks
  20. Google Cloud / DORA (2025). How are developers using AI? Inside our 2025 DORA report
  21. Indeed Hiring Lab (via FRED, St. Louis Fed) (2026). Software Development Job Postings on Indeed in the United States (IHLIDXUSTPSOFTDEVE)
  22. JetBrains (2025). The State of Developer Ecosystem 2025
  23. JetBrains Research (2026). AI Coding Agents: Adoption Trends (Developer Ecosystem Survey 2026)
  24. JetBrains Research (2026). How Much Code Do Developers Really Let Agents Write?
  25. JetBrains Research (2026). Dynamics of Sentiment and Attitudes Toward AI Among Developers
  26. Knowledge at Wharton (Bastani et al., PNAS 2025) (2024). Without Guardrails, Generative AI Can Harm EducationPeer-reviewed version: PNAS 122(26), 2025, doi:10.1073/pnas.2422633122.
  27. METR (2026). We are Changing our Developer Productivity Experiment Design
  28. METR (2025). Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity
  29. Microsoft Research (ACM CHI 2025) (2025). The Impact of Generative AI on Critical Thinking
  30. National Student Clearinghouse (2026). Computer Science Enrollment Is Cooling
  31. NYU (Pearce et al., IEEE S&P 2022) (2021). Asleep at the Keyboard? Assessing the Security of GitHub Copilot's Code Contributions
  32. SignalFire (2025). The SignalFire State of Tech Talent Report — 2025
  33. Stack Overflow (2026). The results of the 2026 Developer Survey are here
  34. Stack Overflow (2026). 2026 Developer Survey — AI
  35. Stack Overflow (2025). 2025 Developer Survey — AI
  36. Stack Overflow (2025). 2025 Developer Survey — Developers (learning to code)
  37. Stack Overflow (2024). 2024 Developer Survey — AI
  38. Stanford (Perry et al., ACM CCS 2023) (2023). Do Users Write More Insecure Code with AI Assistants?
  39. Stanford Digital Economy Lab (2026). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence (Aug 2026 revision)
  40. TechCrunch (2025). A quarter of startups in YC's current cohort have codebases that are almost entirely AI-generated
  41. TechCrunch (2025). GitHub Copilot crosses 20 million all-time users
  42. Veracode (2025). 2025 GenAI Code Security Report (press release)

Cite this page

Free to use with attribution. Copy a citation or an HTML link back to this page.

Citation
Source: Contral, AI Coding Statistics 2026, https://contral.ai/ai-coding-statistics
Reference style
Contral. "AI Coding Statistics 2026." Updated November 2026. https://contral.ai/ai-coding-statistics
HTML link
<a href="https://contral.ai/ai-coding-statistics">AI Coding Statistics 2026</a> (Contral)

AI coding statistics FAQ

What percentage of developers use AI coding tools in 2026?

Nearly all professional developers now use them. JetBrains' Developer Ecosystem Survey 2026 (15,000+ professional developers, May–July 2026) found 90% use AI coding agents at work at least weekly and 68% daily. Stack Overflow's 2025 survey found 84% of respondents using or planning to use AI tools, and Google's 2025 DORA report put AI adoption among software professionals at 90%.

Do AI coding tools make developers faster?

It depends on the task and the developer. Controlled studies range from a 55.8% speedup on a single greenfield task (GitHub Copilot, 2023) and a 26.08% increase in completed tasks across Microsoft, Accenture and a Fortune 100 company, to a 19% slowdown for experienced open-source developers working in their own repositories with early-2025 tools (METR, July 2025). METR's February 2026 follow-up found point estimates favoring AI but with confidence intervals that still include a slowdown.

Is AI-generated code secure?

Not by default. Veracode's 2025 GenAI Code Security Report found AI introduced security vulnerabilities in 45% of 80 curated coding tasks across 100+ LLMs, with a failure rate over 70% for Java. A Stanford user study found developers with an AI assistant wrote significantly less secure code while being more likely to believe it was secure. AI-generated code needs the same or stricter review as human code.

Does using AI hurt learning to code?

It can, depending on how it is used. Anthropic's January 2026 randomized trial of 52 mostly junior engineers found those who learned a new library with AI scored 50% on a comprehension quiz versus 67% for those who coded by hand. Participants who used AI to ask conceptual questions or get explanations scored 65% or higher, while those who delegated code generation scored under 40%.

Is AI reducing junior developer jobs?

Entry-level hiring has been hit hardest, though causality is still debated. Stanford Digital Economy Lab research using ADP payroll data through June 2026 found employment of 22–25-year-olds in AI-exposed occupations is 19% below where it would be had it kept pace with less-exposed peers. SignalFire found new grads were just 7% of Big Tech hires in its 2025 report.

What is vibe coding and how common is it?

Vibe coding means using AI prompted by natural language to write code, often without reading or understanding the result. Collins Dictionary named it Word of the Year 2025. Yet in Stack Overflow's 2025 survey, 72% of developers said vibe coding is not part of their professional work, plus another 5% who said so emphatically.

Can I use these statistics in my article?

Yes. Please cite the original source linked on each stat, and link to this page (https://contral.ai/ai-coding-statistics) as the place you found it. Every stat card has a copy-link button for deep-linking to the exact statistic.

How often is this page updated?

We review the page at least quarterly and whenever a major dataset is released (Stack Overflow Developer Survey, GitHub Octoverse, JetBrains Developer Ecosystem, DORA, and new peer-reviewed studies). The last-updated date at the top of the page reflects the latest full verification pass.