AI Provenance Tracker

Detect AI-generated content with explainable evidence across text, image, audio, and video. Built for fact-checking, newsroom workflows, trust-and-safety teams, and investigation support.

Live Demo | API Docs | Benchmark & Status | Open-Core Model

Start In 3 Steps

1. Upload a file or paste a public URL.
2. Run detection with modality-aware analysis.
3. Review an explainable evidence card (verdict, confidence, timestamp, analysis ID).

Stable vs Experimental

Modality	Status	Primary Entry
Text	Stable	/api/v1/detect/text, /api/v1/detect/url
Image	Stable	/api/v1/detect/image, /api/v1/detect/url
Audio	Experimental	/api/v1/detect/audio
Video	Experimental	/api/v1/detect/video, /api/v1/detect/url

What This Is / Isn't

This is: an evidence-first triage system for authenticity analysis.

This is not: legal proof or a sole-decision engine for high-stakes outcomes.

• Methodology limits: docs/METHODOLOGY_LIMITATIONS.md
• Known operational status: docs/ROADMAP_STATUS.md
• Public/private repo split: docs/OPEN_CORE.md

System Flow

flowchart LR
    A[Input: text, image, audio, video, URL] --> B[Router]
    B --> C[Internal detector by modality]
    C --> D[Calibration and decision band]
    D --> E[Provider consensus]
    E --> F[Evidence card + audit event]
    F --> G[API response + history]

Loading

Fact-Checker Workflow

sequenceDiagram
    participant U as Analyst
    participant W as whoisfake.com
    participant API as API
    participant ENG as Detection Engine
    participant AUD as Audit Store

    U->>W: Upload or paste URL
    W->>API: Detection request
    API->>ENG: Run modality analysis
    ENG-->>API: Verdict + confidence + signals
    API->>AUD: Save analysis_id, timestamp, evidence
    API-->>W: Explainable result card
    W-->>U: Shareable evidence summary

Loading

Evidence Card Example

{
  "analysis_id": "a4f2d7e3-...",
  "content_type": "text",
  "result": {
    "decision_band": "uncertain",
    "is_ai_generated": false,
    "confidence": 0.57,
    "model_version": "text-detector:distilroberta-v1",
    "calibration_version": "calibrated-20260312:general"
  },
  "timestamp": "2026-03-19T12:31:00Z"
}

Quick Start

Docker Compose (recommended)

git clone https://github.com/ogulcanaydogan/AI-Provenance-Tracker.git
cd AI-Provenance-Tracker
cp backend/.env.example backend/.env
make up

• Frontend: http://localhost:3000
• API: http://localhost:8000
• Swagger: http://localhost:8000/docs
• URL analysis UI: http://localhost:3000/detect/url

API Snapshot

URL detection

curl -X POST "https://api.whoisfake.com/api/v1/detect/url" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com"}'

Text detection

curl -X POST "https://api.whoisfake.com/api/v1/detect/text" \
  -H "Content-Type: application/json" \
  -d '{"text":"Your content to analyze"}'

Documentation Map

• API reference: docs/API.md
• Deployment: docs/DEPLOYMENT.md
• Roadmap and run evidence: docs/ROADMAP_STATUS.md
• Open-core boundary: docs/OPEN_CORE.md
• Architecture details: docs/ARCHITECTURE.md

Suggested GitHub About / Topics

About text (recommended): Open-source multimodal AI provenance detection platform with explainable evidence cards, provider consensus, and benchmark-driven quality gates.

Topics (recommended): ai-detection, fact-checking, fastapi, nextjs, provenance, multimodal

Kisa TR Ozet

WhoisFake / AI Provenance Tracker; metin, gorsel, ses ve video iceriklerde AI-uretim sinyallerini aciklanabilir sekilde raporlar.

• Ne yapar: analiz + kanit karti + izlenebilirlik
• Nasil baslanir: URL yapistir veya dosya yukle, sonucu paylas
• Sinirlar: olasiliksal sonuc uretir; tek basina hukuki kanit degildir

License

MIT - see LICENSE.