Ai Ranko
Ai RankoToolsVibe Code Detector
Probabilistic forensic analysis · 90+ signals · Zero fake scores

Vibe Code Detector

Estimate the probability that a website is Human Coded, AI / Vibe Coded, CMS Generated, or a Hybrid of the two. This is a deep forensic analysis — hosting, images, HTML/CSS/JS patterns, packages, content, and dozens more signals each cast a weighted vote — not a certainty claim that any specific AI tool was used.

90+ signals
CMS-aware
Weighted probability
JSON export

Public URL only. We fetch via our server (no CORS), follow redirects, and parse raw HTML + response headers.

Analysis takes 15–30 seconds. Weighted probabilistic scoring across 90+ signals — no guesswork.

How the Vibe Code Detector Works

Three inputs, eight real audits, one honest verdict on whether your site is vibe-coded.

01

Provide Source

Paste a URL, GitHub repo URL, or your raw code. URL mode fetches HTML + HTTP headers, GitHub mode pulls up to 30 prioritized source files, paste mode analyzes verbatim.

02

Real Vibe-Code Audit

We count actual patterns linked to AI tools like v0, Bolt, Lovable, Cursor, Windsurf and Replit: data-testid attributes, AI-style "// Here we…" comments, overly-descriptive variable names, generator meta tags, Tailwind utility soup and more. Every number is a real count.

03

Vibe Score Report

Eight per-analyzer cards with score gauges, exact counts, color palettes, heading trees, sentence distribution, security tables, image audits and metadata checklists. Export the full report as JSON.

8 Real Vibe-Code Analyzers

Each one counts something real in your source. No randomization, no hidden estimates, no LLM guessing.

Inline Code

Counts inline scripts, style attributes, event handlers, HTML comments, AI-style comments inside scripts, and data-testid attributes (a strong vibe-code signal).

CSS Quality

Tailwind utility soup (elements with 8+ classes), !important abuse, generic CSS variable names, long class names, and fonts — the real fingerprints of v0/Lovable output.

HTML Structure

Semantic tag percentage, deep div nesting (>6 levels), wrapper div ratio, empty elements, missing landmarks, broken heading hierarchy.

JavaScript

Generic variable names, AI-style "// Here we…" comments, overly-descriptive identifiers (>25 chars), leftover console.logs, TODO/FIXME, magic numbers, async density.

Copywriting

AI buzzwords (seamless, leverage, delve, robust, harness…), filler phrases, vague CTAs, low sentence-length variance — the verbal fingerprints of ChatGPT/Claude copy.

Security Headers

URL-mode only. Inspects CSP, X-Frame-Options, HSTS, Referrer-Policy, Permissions-Policy and more — vibe-coded sites almost always ship with zero security headers.

Image Quality

Missing alt, generic alt, dimensions, Next.js Image usage, and AI-typical filenames like "placeholder.jpg" or "hero-image.png".

Metadata & SEO

Title length, description, generator tag (huge red flag — explicitly detects v0, Bolt, Lovable, Cursor, Replit), Open Graph, Twitter, canonical, robots, JSON-LD.

Three Ways to Detect Vibe Code

Pick the input that matches your audit goal.

Website URL

Fetched server-side with a realistic User-Agent. Follows redirects. Gets HTML + HTTP security headers — the fastest way to check if a live site is vibe-coded.

Best for: checking if a competitor or client site was built with v0, Bolt, or Lovable.

GitHub Repo

Pulls up to 30 source files prioritized by location (root, src/, app/, components/). Handles any default branch automatically.

Best for: auditing your own repo or any public repo for vibe-code patterns.

Paste Code

Analyzes any HTML, CSS, JavaScript, JSX, or mixed source you paste — minimum 100 characters.

Best for: a quick "is this vibe-coded?" check on a snippet.

What Is “Vibe Coding” — And Why Detect It?

Vibe coding is the 2025 term for shipping software written end-to-end by AI tools — v0, Bolt, Lovable, Cursor, Windsurf, Replit Agent, Devin — with little or no human refactoring. The output works, but it leaves a fingerprint: utility-class soup, deep wrapper divs, “// Here we handle…” comments, generic naming and missing security headers.

Most AI detectors output one black-box probability. The Vibe Code Detector does the opposite — every number traces back to a count, a regex match or an HTTP header you can read in the report and audit yourself.

  • Every score is computed from real counts — no model probabilities, no random seeds
  • Each analyzer card shows the raw measurements that drove its score
  • Re-running on the same input always returns the same Vibe Score
  • Full JSON export gives you the entire decision trail
  • Severity bands map to fixed score ranges, not guesses
  • Explicitly detects v0, Bolt, Lovable, Cursor, Windsurf and Replit generator tags
8Real auditsevery metric counted
30Files per repoGitHub mode
11AI tools detectedv0, Bolt, Lovable…
0Fake scoreszero estimation

Frequently Asked Questions

What does "vibe code" mean?

Vibe coding is the 2025 term for software shipped largely or entirely from AI coding tools — v0, Bolt, Lovable, Cursor, Windsurf, Replit Agent, GitHub Copilot, Devin, ChatGPT or Claude — without significant human refactor. The Vibe Code Detector looks for the visual and structural fingerprints these tools leave behind.

How does the Vibe Code Detector work?

It fetches your source (HTML from a URL, files from a GitHub repo, or your pasted code) and runs 8 deterministic audits that count real patterns. Every score is calculated from real counts — no AI guessing, no randomization. The minimum analysis window is 10 seconds so the result reflects real measurement, not a coin flip.

What does the Vibe Score actually mean?

Higher = more vibe-code patterns detected. Lower = more hand-crafted. The bands: Likely human-written (0–24), Mixed signals (25–49), Likely vibe-coded (50–74), Strongly vibe-coded (75–100).

Which AI coding tools can it detect?

The metadata analyzer explicitly flags generator tags from v0 (Vercel), Bolt, Lovable, Cursor, Windsurf, Replit, Devin, GitHub Copilot, ChatGPT and Claude. The other 7 analyzers detect generic AI-coding fingerprints regardless of which tool produced them.

Do you store my data?

No. The tool fetches your URL or repo server-side and analyzes the response in memory. Nothing is logged or stored.

What GitHub repos can I analyze?

Any public repository. Private repos require auth which the tool does not currently support. The unauthenticated GitHub API has a 60-requests-per-hour limit, so heavy use may briefly rate-limit.

Why does Security Headers only run for URLs?

Security headers are HTTP response headers — they only exist when a live server responds. Pasted code and GitHub files have no associated HTTP response. Worth noting: vibe-coded sites almost always ship with zero security headers, so URL mode catches one of the strongest signals.

Can the Vibe Code Detector be fooled?

Yes. The detector measures patterns commonly associated with AI-generated code, but a skilled developer can write code matching those patterns, and AI tools can be prompted to avoid them. Treat the Vibe Score as a strong signal, not a verdict.

Is it really free?

Yes — no sign-up, no watermark, no rate limits beyond what the underlying services impose.

Is It Vibe-Coded? Find Out Now.

Paste a URL, a GitHub repo, or your raw code — and get a real, measurement-based Vibe Score report in under a minute.

Site Audit Tool AI Content Detector AI Content Optimizer