Claim-level AI trust verification. Verify AI responses with real evidence.
AI can sound confident even when it's wrong. Hallucination Hunter splits AI responses into individual factual claims, searches real web sources for evidence, verifies each claim, and gives you a transparent Trust Score — backed by real sources.
AI Response
→ Claim Extraction (split into individual claims)
→ Evidence Retrieval (real web search via Tavily)
→ Claim Verification (LLM verifies using ONLY retrieved evidence)
→ Confidence Scoring (source quality, count, agreement)
→ Trust Score (weighted by importance + severity)
→ VERIFIED / FALSE / UNVERIFIABLE per claim
→ Source attribution + Corrections
| Feature | Status |
|---|---|
| Paste AI response | ✅ |
| Share-to-Verify (Android) | ✅ |
| Screenshot OCR | ✅ |
| Document upload (PDF/DOCX/TXT) | ✅ |
| Claim extraction | ✅ |
| Real evidence retrieval (Tavily) | ✅ |
| Source quality scoring | ✅ |
| Per-claim verdict (VERIFIED/FALSE/UNVERIFIABLE) | ✅ |
| Confidence score | ✅ |
| Trust Score | ✅ |
| Correction generation | ✅ |
| Verified Answer rewrite | ✅ |
| SSE real-time progress | ✅ |
| Verification history | ✅ |
| Guest mode | ✅ |
| Agent API for external AI | ✅ |
| Dark-mode professional UI | ✅ |
| Layer | Technology |
|---|---|
| Mobile | React Native + Expo (TypeScript) |
| Backend | Python FastAPI |
| AI / LLM | Google Gemini 1.5 Flash |
| Evidence | Tavily Search API |
| Database | PostgreSQL |
| Cache | Redis |
| OCR | Tesseract (pytesseract) |
| Auth | JWT (optional, guest-first) |
| Streaming | Server-Sent Events (SSE) |
See SETUP.md for full installation instructions.
# 1. Clone
git clone <repo>
cd hallucination-hunter
# 2. Configure
cp .env.example .env
# Edit .env with your API keys
# 3. Start database + Redis
docker-compose up postgres redis -d
# 4. Start backend
cd backend && pip install -r requirements.txt
uvicorn app.main:app --reload
# 5. Start mobile app
cd mobile && npm install
npx expo start --androidhallucination-hunter/
├── mobile/ # React Native + Expo (TypeScript)
├── backend/ # Python FastAPI
├── agent/ # Modular verification agent (11 tools)
├── database/ # PostgreSQL migrations
├── tests/ # pytest + Jest tests
└── docs/ # Documentation
- Gemini API (free tier): https://aistudio.google.com
- Tavily Search API (free tier: 1000 searches/month): https://tavily.com
"Hallucination Hunter is a mobile AI trust layer that verifies AI-generated information at the individual claim level, retrieves supporting evidence, detects contradictions, assigns confidence scores, and gives users transparent source-backed results before they trust or share AI-generated information."
Core differentiator: CLAIM-LEVEL AI TRUST VERIFICATION
MIT