Pro & Business Plans
Enhanced Grounding
Chunk-level search with AI reranking and confidence scoring. The difference between GPT content and truly sourced content.
Standard vs Enhanced Grounding
Standard Search
Free & Starter Plans
- Document-level search
- 3-5 sources per article
- Basic claim validation
Enhanced Grounding
Pro & Business Plans
- Chunk-level search (13,000+ passages)
- Hybrid Search: BM25 + semantic + RRF
- LLM Reranking with GPT-4o-mini
- 5-10 verified sources per article
- Confidence scoring (high/medium/low)
How does it work?
Production-grade RAG architecture
Hybrid Search
Combines full-text (BM25) and semantic (KNN) search with Reciprocal Rank Fusion (K=60) for more relevant results.
LLM Reranking
GPT-4o-mini evaluates each passage on a 0-10 scale. Final score: 30% original + 70% reranking score.
Semantic Chunking
Documents split into 6000-character chunks with 200 overlap. Each chunk is vectorized individually.
Confidence Scoring
Confidence score based on: source coverage, semantic similarity, diversity. Levels: high, medium, low, insufficient.
Ready for truly sourced content?
Upgrade to Pro or Business plan to activate Enhanced Grounding on all your articles.