⚡ VibeCheck — Vibe Coding Detector

AI-powered app that analyzes conversations and determines the vibe or sentiment using modern AI models.

🛠 Tech Stack

Frontend

• Next.js

• React

• Tailwind CSS

Backend / APIs

• Serverless API routes

• AI model integration

Infrastructure

• Vercel (hosting & deployment)

• GitHub (source control)

📸 Screenshots

Paste any URL. Find out if the site was vibe coded.

📖 Read the full blog post →

The Problem

"Vibe coding" — building entire websites by prompting AI tools like v0, Bolt.new, Lovable, Cursor, and others — has exploded in popularity. While this unlocks incredible speed, it also introduces real concerns:

• Quality & Maintainability: Vibe-coded sites often ship with deeply nested DOM trees, generic placeholder copy, unused boilerplate, and cookie-cutter layouts that become technical debt.
• Authenticity & Trust: Clients, employers, and users increasingly want to know whether a site was genuinely crafted or rapidly generated. Portfolios filled with AI output misrepresent skill.
• Security & Best Practices: Auto-generated code frequently skips fundamentals — proper SEO metadata, robots.txt, accessibility, error handling — creating hidden risks.
• Hiring & Evaluation: When reviewing developer portfolios or freelance deliverables, there's no easy way to distinguish handcrafted work from AI-assembled output.

There was no tool to answer the simple question: "Was this website vibe coded?"

The Solution

VibeCheck is a web-based analysis engine that scans any public URL and produces a Vibe Score (0–100) with a detailed breakdown across 7 detection categories. It uses heuristic pattern matching — no AI/LLM is used in the analysis itself — to look for the telltale fingerprints that AI-assisted code generators leave behind.

Detection Categories

Category	What It Checks	Weight
🤖 AI Platform Signatures	v0.dev data attributes, Bolt.new markers, Lovable/GPTEngineer references, Replit domains, meta generator tags	30%
🧩 UI Library Patterns	shadcn/ui class combos, Radix UI data-radix-* attributes, shadcn CSS variables (--radius, --primary), Lucide icons	15%
⚡ Framework Detection	Next.js (__next, _next/static), Vite module scripts, React root markers, Nuxt, Astro, SvelteKit	10%
📝 Content Signals	Generic marketing phrases ("Transform your…", "Revolutionize…"), stock image CDNs, placeholder text, generic CTAs	15%
🔬 Code Style Analysis	DOM nesting depth, Tailwind utility-class density, AI-style HTML comments, inline event handlers	10%
🚀 Deployment Signals	Vercel/Netlify/Railway headers & domains, missing robots.txt, missing sitemap.xml	10%
🎨 Design Patterns	Glassmorphism (backdrop-blur), gradient overuse, cookie-cutter section layouts, card grids, neon/glow effects	10%

Verdict Scale

Score	Verdict
🟢 0–25	Likely Human-Crafted
🟡 26–50	Mixed Signals
🟠 51–75	Probably Vibe Coded
🔴 76–100	Almost Certainly Vibe Coded

Each finding includes a confidence level (High / Medium / Low) and supporting evidence, so you can judge for yourself.

How to Run

Prerequisites

• Python 3.9+ installed on your machine
• pip (comes with Python)

1. Clone the Repository

git clone https://github.com/ashish-jabble/vibe-check.git
cd vibe-check

2. Create a Virtual Environment

python3 -m venv venv
source venv/bin/activate        # macOS / Linux
# venv\Scripts\activate          # Windows

3. Install Dependencies

pip install -r requirements.txt

4. Start the Server

python app.py

The app will be available at http://127.0.0.1:5000.

5. Analyze a Site

1. Open http://127.0.0.1:5000 in your browser.
2. Paste any public URL into the input field.
3. Click Analyze and wait a few seconds.
4. Review the Vibe Score and detailed category breakdown.

Project Structure

vibe-check/
├── app.py                  # Flask server — serves UI & /api/analyze endpoint
├── analyzer.py             # Core detection engine — 7 category heuristics
├── requirements.txt        # Python dependencies
├── .gitignore              # Git ignore rules
├── README.md               # You are here
├── static/
│   ├── css/
│   │   └── style.css       # Premium dark UI with animations
│   └── js/
│       └── app.js          # Frontend logic — form, loading, results
└── templates/
    └── index.html          # Main HTML template

API Usage

You can also call the analysis engine directly via the API:

curl -X POST http://127.0.0.1:5000/api/analyze \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com"}'

Response:

{
  "url": "https://example.com",
  "overall_score": 42,
  "verdict": "Mixed Signals",
  "verdict_emoji": "🟡",
  "categories": {
    "ai_platforms": { "name": "AI Platform Signatures", "score": 0, "findings": [] },
    "ui_libraries": { "name": "UI Library Patterns", "score": 36, "findings": [...] },
    ...
  }
}

Limitations

• Heuristic-based: Results are indicative, not definitive. A high score doesn't guarantee AI generation; a low score doesn't guarantee human authorship.
• Static analysis only: The tool fetches raw HTML via requests — it does not execute JavaScript. SPAs that rely heavily on client-side rendering may show fewer signals.
• False positives: Legitimate sites using shadcn/ui, TailwindCSS, or Vercel may trigger some signals. The weighted scoring system mitigates this.
• Evolving landscape: As AI tools mature, their output patterns will change. The heuristics will need regular updates to stay relevant.

Contributing

Contributions are welcome! Some ideas:

• Add new heuristics for emerging AI tools (Windsurf, Replit Agent, etc.)
• Improve scoring weights based on empirical testing
• Add JavaScript rendering support (Playwright/Selenium)
• Build a browser extension for inline analysis
• Create a public API with rate limiting

License

MIT — use it however you like.

⭐ Support

If you like this project, please consider starring the repository on GitHub.