AI search for websites
AI search works out what a query means and finds the right content, even when the wording differs.
Background
Classic site search only matches the exact words in the text. Someone searching for ‘cancel contract’ will not find a page called ‘Ending your agreement’. The outcome is empty results pages and extra support requests.
Semantic and hybrid search
Semantic search turns content and queries into vectors (embeddings) that capture meaning, so it also finds synonyms and paraphrases. Paired with keyword search, it becomes hybrid search: exact matches for product names and item numbers, relevant content for questions. Typo tolerance and filters are included.
AI answers from your content
Retrieval-augmented generation (RAG) can add an AI answer above the results. It summarises the relevant pages in a few sentences and links to its sources. When nothing fits, it says so plainly.
Technology
Depending on scope, search runs on Algolia, Meilisearch or Typesense, or on OpenSearch and pgvector within your own infrastructure. It connects to Statamic, Shopify or Laravel applications, and the index updates whenever content changes.
Self-hosted, all data stays within your infrastructure.
Measuring success
We track searches with no results, clicks on results and abandoned searches, which also reveals missing content. Before launch, old and new search are compared against real queries.