How Does a Search Engine Work? From a Query to the Right Answer in Milliseconds

Every day we type a few words into a search box and expect the right answer back instantly, out of billions of documents. But what actually happens in those milliseconds? In this talk, we open up the black box and walk through the core concepts that power every search engine, from Google to enterprise AI search.

Try the Demo

Try out the search engine from the talk yourself: viscon-search.lukasre.ch

The Journey of a Query

We follow a single query on its journey through a search engine, from the moment you hit enter to the ranked results on your screen.

Crawling and ingestion

How a search engine discovers and collects the documents it can search over

Indexing

The inverted index, the data structure that makes searching billions of documents in milliseconds possible

Tokenization and text processing

How raw text becomes something a machine can match against

Retrieval

Finding the candidate documents that could answer a query

Ranking

The heart of search: deciding which results actually go on top, and why

Key Concepts

• Why keyword search (BM25/TF-IDF) and semantic search (embeddings, vector search) solve different problems, and why modern systems combine both
• What embeddings actually are and how "meaning" becomes math
• The difference between retrieval (finding candidates) and ranking (ordering them well)
• How relevance is measured and why "the right answer" is harder to define than it sounds
• Where Large Language Models fit into the modern search stack, and how RAG turns search into answers

Build One Yourself

Want to go hands-on? In the Build Your Own RAG Pipeline workshop, you build a complete retrieval pipeline that answers questions with cited sources.