A cross-browser extension + local backend for detecting AI-generated images, videos, and articles — with built-in fact-checking powered by Google Gemini and web search grounding.

AI-generated media is increasingly realistic and widespread. Slop-Detector gives users instant, on-demand detection scores and fact-checks for any image, video, or article — right from their browser. It combines the AI-or-Not API for confidence scoring with Gemini-powered analysis for deeper inspection, running everything through a local FastAPI backend so API keys never touch the browser.

Target users: Journalists, researchers, educators, fact-checkers, and anyone who wants to verify the authenticity of online media.

• Right-click any image to scan it instantly via context menu
• Page picker — interactively select any image with a visual overlay
• File upload — upload an image from your device
• Batch scan — scan every image on the current page at once with configurable sensitivity
• Auto-scan — optionally scan images automatically on page load
• Inline badges — color-coded confidence badges appear directly on the page

• URL analysis — paste any YouTube, TikTok, or direct video URL
• File upload — upload a video from your device (max 200 MB, 30 seconds)
• Gemini analysis — AI detection + structured fact-check of key claims with verdicts
• Too-long handling — videos over 60 seconds return a download token for a trimmed clip

• Grab current page — extracts article text from the active tab automatically
• Paste text — paste any article, headline, or content directly
• AI detection — score (0–100), writing pattern signals, and plain-English summary
• Fact-check — extracts claims, verifies them via Gemini with Google Search grounding, returns a verdict and source articles

• Scan history — up to 50 past results with thumbnails, scores, and timestamps
• Result caching — 7-day TTL per URL so repeated scans are instant
• Backend fallback — if the local backend is unreachable, falls back directly to the AI-or-Not API
• Connection indicator — live status dot shows whether the backend is reachable

AI_Slop_Detector/
├── backend/
│   ├── main.py                 # FastAPI app, CORS, health endpoint
│   ├── helper.py               # API key loading from apikeys.env
│   ├── apikeys.env             # API keys (not committed)
│   ├── pyproject.toml          # Python dependencies (uv)
│   ├── routes/
│   │   ├── image.py            # /image endpoints
│   │   ├── video.py            # /video endpoints
│   │   ├── text.py             # /text endpoints
│   │   └── factcheck.py        # /factcheck endpoints
│   └── services/
│       ├── image_service.py    # AI-or-Not + Gemini image logic
│       ├── video_service.py    # AI-or-Not + Gemini video logic
│       ├── text_service.py     # Gemini text AI detection
│       └── factcheck_service.py# Gemini claim extraction + web-grounded fact-check
└── frontend/
    ├── manifest.json           # MV3 extension manifest
    ├── background/
    │   └── background.js       # Service worker, message bus, context menu
    ├── content/
    │   ├── content.js          # Page badges, element picker, image collection
    │   └── content.css         # Badge and overlay styles
    ├── popup/
    │   ├── popup.html          # Popup UI — Image, Video, Article tabs
    │   ├── popup.js            # Tab logic, API calls, result rendering
    │   └── popup.css           # Popup styles with dark mode support
    └── utils/
        └── api.js              # Fetch utilities, score normalization, caching

• Python 3.10+
• uv (or pip)
• Chrome or Firefox
• AI-or-Not API key
• Google Gemini API key

cd backend
uv venv
source .venv/bin/activate        # Windows: .venv\Scripts\activate
uv sync

Create backend/apikeys.env:

AIORNOT_KEY=your_aiornot_api_key
GEMINI_API_KEY=your_gemini_api_key

Start the server:

uvicorn main:app --reload

Confirm it's running: http://localhost:8000/health

Optional: Install ffprobe (part of FFmpeg) to enable video duration enforcement on uploads. Install yt-dlp for downloading YouTube/TikTok/Instagram videos for fact-checking.

Chrome / Chromium:

1. Go to chrome://extensions
2. Enable Developer mode
3. Click Load unpacked and select the frontend/ folder

Firefox:

1. Go to about:debugging#/runtime/this-firefox
2. Click Load Temporary Add-on
3. Select frontend/manifest.json

Open the extension popup. Settings are accessible via the browser's extension settings or the Settings tab (if present):

Keyboard shortcut: Ctrl+Shift+U / Cmd+Shift+U — open the popup

{
  "verdict": "likely_real | uncertain | likely_ai",
  "ai_score": 42,
  "ai_signals": ["no named sources", "generic list structure"],
  "claims": [{"text": "...", "assessment": "supported | contradicted | unverifiable", "explanation": "..."}],
  "summary": "Plain English summary of findings."
}

{
  "claims": ["Claim one", "Claim two"],
  "factuality_score": 75,
  "verdict": "Mostly True",
  "explanation": "2–3 sentence summary.",
  "articles": [{"title": "...", "url": "...", "snippet": ""}]
}

Scores represent the likelihood the content is real (not AI-generated):

For fact-check verdicts: