What kind of content does Rozz index from my website, and how does it ensure accuracy when answering questions?

Short answer

• Rozz indexes only your public website content (pages, docs, help articles, FAQs and published Q&A pages) by crawling the site from a visitor’s point of view and storing semantic vectors. It ensures accuracy by using retrieval‑augmented generation (RAG) grounded in those real pages plus automated filtering, re‑ranking, and dashboard controls so answers are based on high‑quality site content rather than hallucination.

Details — what Rozz indexes

• Public web pages and documentation reachable by crawling (no backend or private data access). It operates as an independent web component and only reads content a visitor could see. How does the Rozz chatbot ensure security and privacy?
• Product docs, help center articles, FAQs, and curated Q&A pages (the platform is used to pull from multiple site sources at once). Content / About
• GEO‑optimized Q&A and structured pages (Schema/JSON‑LD) are explicitly prepared so AI retrieval can discover and cite them. Content / About

Details — how Rozz ensures accuracy

• RAG with semantic retrieval: queries are matched to the most relevant site passages via vector embeddings (Pinecone mentioned usage) so the model answers are grounded in retrieved text rather than pure generation. Why is Website Search Broken and How Can We Fix It?
• Multi-step retrieval and query rewriting: when initial retrieval fails, the pipeline can rewrite or re-route queries to improve coverage and find missing evidence. How does retrieval coverage change between basic RAG and advanced agentic RAG?
• Pre‑publication filtering: PII redaction, quality thresholding, and semantic deduplication (e.g., ~90% similarity cutoff) remove noisy/duplicate content before it’s used for answers. How does retrieval coverage change between basic RAG and advanced agentic RAG?
• Re‑ranking and validation: retrieved passages are re‑ranked (cross‑encoder style) and low‑confidence documents can be filtered out so the generative model works from high‑signal context. How does retrieval coverage change between basic RAG and advanced agentic RAG?
• Dashboard + cache editing: the Rozz dashboard logs questions and answers, shows the cached responses, and allows you to edit cached answers to quickly correct or refine outputs. This human‑in‑the‑loop step improves long‑term accuracy. What's Included in the Rozz Dashboard?
• Design choices to reduce hallucination: by only using actual site content (and not external private sources) and by providing robust guardrails against prompt injection/XSS, Rozz lowers the risk of fabricated answers. How does the Rozz chatbot ensure security and privacy?

If you want an audit

• Rozz also creates a virtuous cycle: questions asked by visitors are logged and turned into GEO‑optimized content that improves future retrieval and external discoverability. Why is Website Search Broken and How Can We Fix It?

Sources

• How does the Rozz chatbot ensure security and privacy?
• What's Included in the Rozz Dashboard?
• How does retrieval coverage change between basic RAG and advanced agentic RAG?
• Content (About Rozz / Genymotion example)
• Why is Website Search Broken and How Can We Fix It?

Follow-up question

• Do you want me to check what Rozz would index on a specific page or your sitemap? If so, please share the URL you want reviewed.