What are the key steps to designing an AI search engine that cites sources and ranks answers by confidence rather than just relevance?

1. Understand the complexity beyond simple retrieval and generation. 2. Retrieve sources with diversity using Maximum Marginal Relevance. 3. Extract discrete claims from each source. 4. Detect contradictions among claims. 5. Build a confidence scoring system based on agreement and source authority. 6. Generate answers with citations and acknowledgment of uncertainty. 7. Rank answers by confidence, not popularity.