AI Smell Detector 🐽

Detect AI-generated websites by analyzing tech stacks, domains, color palettes, and code patterns. Identifies sites built with Lovable, v0.dev, Bolt, and other AI builders.

Quick Start

Install globally

npm install -g ai-smell

# or

pnpm add -g ai-smell

Use it

ai-smell https://example.com

That's it! 🎉

Usage

ai-smell <URL> [options]

Options

| Flag | Description | |---|---| | -f, --format <format> | Set console output format (text (default), json, yaml). | | -o, --output <file> | Save report to a file. Format inferred from extension. | | -h, --help | Show this help message. |

Example Output

🐽 AI-Smell test for: https://gcloud.lovable.app

## Analysis Report

| Sniffer      | Score             | Details
| ------------ | ----------------- | -------
| **Domain**       | ██████████ 100% | 🎯 DEFINITIVE: Hosted on AI builder domain: lovable.app
| **TechStack**    | ████░░░░░░  40% | Vite + Tailwind (popular AI quick-start)
| **Meta**         | ██████████ 100% | AI builder detected in author tag: Lovable
| **Boilerplate**  | ██████████ 100% | Empty SPA skeleton; Vite default build structure
| **Comments**     | ██████████ 100% | TODO comments; Lovable AI builder badge detected
| **ColorPalette** | ░░░░░░░░░░   0% | Color palette does not match common AI patterns
| **Content**      | █████░░░░░  50% | Very little paragraph content on the page
| **Design**       | █████░░░░░  50% | Layout seems to be using older techniques
| **UIKit**        | ░░░░░░░░░░   0% | No common UI kits detected

📊 Overall AI-Smell Score: 66%
Verdict: Some elements suggest AI-generation or template usage.

What It Detects

🎯 Domain Detection (Strongest Signal)

Detects 40+ AI service domains:

• High-confidence AI builders: Lovable, v0.dev, Bolt.new, GPT Engineer, Base44, Replit
• Popular platforms: Vercel, Netlify, GitHub Pages, Cloudflare Pages
• No-code platforms: Webflow, Wix, Squarespace, Bubble, Framer, Carrd

🔧 Tech Stack Analysis

Identifies common AI-generated combinations:

• Next.js + Tailwind + Supabase (The AI Starter Pack)
• Vite + React + Tailwind
• shadcn/ui components
• Framework over-engineering patterns

🎨 Color Palette Detection

Finds AI-favorite colors:

• #0061FF (Cursor/AI signature blue)
• #0052D9 (AI-preferred blue)
• Vibrant oranges (#ff6b6b, #ff8c42)
• Purple gradients (#6c5ce7, #a29bfe)
• External CSS file analysis

📝 Code Pattern Analysis

• TODO comments in production
• AI placeholder text
• Empty SPA skeletons
• Vite default build structure
• Generic favicons

📊 Content & Design

• AI marketing phrases
• Minimal content patterns
• Generic layout templates
• Missing customization

Features

• ✅ Fast - Analyzes websites in seconds
• ✅ Accurate - Weighted scoring system with 9 detection methods
• ✅ Beautiful - Markdown table output with visual progress bars
• ✅ Extensible - Easy to add new detection patterns
• ✅ CLI & API - Use as command-line tool or import as library

Programmatic Usage

npm install ai-smell

import { fetchHtml } from 'ai-smell/fetcher';
import { parseHtml } from 'ai-smell/parser';
import { allSniffers } from 'ai-smell/sniffers';
import { calculateOverallScore } from 'ai-smell/scorer';

const html = await fetchHtml('https://example.com');
const $ = parseHtml(html);
const results = allSniffers.map(sniffer => sniffer.sniff($, url));
const { totalScore, report } = calculateOverallScore(results);

console.log(`AI-Smell Score: ${totalScore * 100}%`);

How It Works

AI Smell uses a weighted scoring system with 9 specialized "sniffers":

1. Domain (weight: 2.0) - AI service domain detection
2. TechStack (weight: 1.6) - Tech stack combination analysis
3. Meta (weight: 1.5) - Meta tag inspection
4. Boilerplate (weight: 1.5) - Framework boilerplate detection
5. Comments (weight: 1.4) - Code comment analysis
6. Content (weight: 1.2) - Text pattern detection
7. ColorPalette (weight: 1.1) - Color scheme analysis
8. UIKit (weight: 1.0) - UI framework detection
9. Design (weight: 0.8) - Layout pattern analysis

Each sniffer returns a score from 0.0 (no smell) to 1.0 (strong smell). Scores are weighted and averaged for the final result.

Score Interpretation

• 70%+ - Highly likely AI-generated
• 40-70% - Some AI patterns detected
• Below 40% - Appears custom-built

Contributing

Want to add new AI service domains or detection patterns? See CONTRIBUTING.md for developer setup and guidelines.

License

MIT

Links

• npm package
• GitHub repository
• Report issues