Methodology

Last Updated: August 7, 2026

How We Research Vibe Coding Statistics

Every statistic published on Vibecodingstats.dev is selected, verified, and reviewed before it appears on our website. Our goal is to provide readers with accurate, current, and transparent data on AI-assisted coding, developer adoption, platform economics, and code quality that can be confidently referenced in research, journalism, investment analysis, and business decisions.

Rather than repeating widely circulated numbers, we trace each statistic back to its original publication whenever possible.

Our Research Process

1. Source Discovery

We continuously monitor new research on vibe coding, AI-assisted development, and the broader developer tooling ecosystem from technology platforms, research firms, security organizations, universities, and industry publications.

Typical sources include:

  • Stack Overflow (Developer Survey)
  • GitHub (Octoverse, Copilot disclosures)
  • Vercel (Developer Survey)
  • Anthropic (Claude Code disclosures)
  • Cursor / Anysphere
  • Lovable (Build Economy Report)
  • Replit
  • McKinsey
  • Gartner
  • Forrester
  • Grand View Research
  • Veracode
  • CodeRabbit
  • GitClear
  • METR
  • Uplevel
  • Stanford University and ACM research
  • Georgia Tech (Vibe Security Radar)
  • Escape.tech and OX Security
  • OWASP
  • Cloud Security Alliance
  • Bloomberg and TechCrunch (funding and revenue reporting)
  • arXiv (when relevant)

2. Primary Source Verification

Before publishing a statistic, we verify:

  • Original publication
  • Publication date
  • Research methodology
  • Sample size (when available)
  • Geographic and industry coverage
  • Whether the figure refers to vibe coding (prompt-to-app) or the broader AI-assisted coding category
  • Whether newer data exists

Whenever possible, we cite the original source rather than secondary articles that reference it.

3. Cross-Checking

If multiple organizations report the same metric differently, for example adoption rates or platform ARR, we compare:

  • Publication dates
  • Methodology (self-reported vs measured, survey vs telemetry)
  • Definitions (weekly use, daily use, “using or planning to use”)
  • Scope (US vs global, professional vs total developers)
  • Survey population and sample size
  • Whether figures are disclosed, estimated, or projected

When meaningful differences exist, we either explain them side by side or omit the statistic rather than present potentially misleading data.

4. Reconciling Reported and Measured Data

Vibe coding is unusual in that self-reported productivity often diverges sharply from measured productivity. GitHub reports 55% faster task completion, while METR measured developers 19% slower in legacy codebases despite feeling 20% faster.

When this pattern appears, we:

  • Publish both figures rather than choosing the more favorable one
  • Label each as reported or measured
  • Note the study design (survey, controlled trial, telemetry analysis)
  • Explain the gap where it materially affects interpretation

5. Editorial Review

Every report goes through multiple editorial checks to ensure:

  • Numerical accuracy
  • Consistent formatting across tables and charts
  • Correct attribution to primary sources
  • Updated dates
  • Working citations
  • Clear distinctions between confirmed, estimated, and projected figures
  • Clear explanations of context

No article is published until it passes our review process.

How We Handle Conflicting Statistics

Vibe coding is one of the fastest-changing categories in software. Different organizations often report different values because they use different methodologies, define the market differently, or cover different time periods. Market size figures for 2026, for example, range from $4.7 billion (vibe coding segment only) to $10 billion (all AI code tools).

When this happens, we:

  • Prefer primary research over aggregated summaries
  • Explain differences in scope or definition where appropriate
  • Avoid averaging figures that measure different things
  • Distinguish between the prompt-to-app segment and the wider AI-assisted coding market
  • Label estimates, projections, and analyst forecasts as such
  • Exclude unsupported figures

How We Handle Private Company Financials

Revenue and valuation figures for private companies such as Cursor, Lovable, Replit, and Base44 are based on media reports, investor disclosures, and analyst estimates, not audited financials.

When reporting these figures, we:

  • Cite the reporting outlet and disclosure date
  • Distinguish confirmed ARR from run-rate and estimated figures
  • Update numbers as new funding rounds or earnings reports are published
  • Remove figures that are contradicted by newer, more credible reporting

How Often We Update Statistics

Our reports are reviewed regularly throughout the year, with priority given to high-velocity metrics such as platform ARR, funding rounds, and CVE counts.

Updates may include:

  • Newly published developer surveys
  • Revised market forecasts
  • Platform-reported ARR, user counts, and valuations
  • New security research and vulnerability studies
  • Updated CVE tracking from Georgia Tech and others
  • Regulatory developments (EU AI Act, App Store policies)
  • Corrections issued by original publishers

Each report displays its latest update date.

Our Coverage

Our research covers statistics across the vibe coding and AI-assisted development ecosystem, including:

  • Vibe coding adoption
  • AI-assisted coding tools
  • Developer productivity
  • Code quality and technical debt
  • AI code security and vulnerabilities
  • Platform market share and ARR
  • Funding and valuations
  • Enterprise and Fortune 500 adoption
  • Non-developer and citizen developer adoption
  • Startup and YC cohort trends
  • Prompt-to-app platforms
  • Agentic coding and autonomous engineering
  • Developer trust and sentiment
  • Governance and regulatory trends
  • Market size and forecasts

Transparency

If we cannot verify a statistic through a credible source, we do not publish it.

If a figure is contested, we present the range and its sources rather than picking a single number. If an error is identified after publication, we review the evidence and correct the page as quickly as possible, noting significant corrections at the bottom of the affected page.

For corrections, data submissions, or methodology questions, contact the editorial team at vibecodingstats.dev.

Scroll to Top