Search and discovery

How help center search ranks articles with keyword and semantic matching, where search and Ask AI appear, and how to write articles that get found.

Search is how most visitors find answers in your help center. This article explains how search ranks articles, where it appears, how Ask AI works, and how to write articles that get found.

How search works

Help center search combines two methods:

Keyword search (40% of the score): full-text matching on article titles and content. Fast and precise for exact terms, and works with no AI configured.
Semantic search (60% of the score): matches by meaning using vector embeddings. A visitor searching "can't log in" can find an article titled "Reset your password" even though the words differ.

The two scores are combined to rank results.

Note: On Quackback Cloud, semantic search works out of the box. A self-hosted install needs an embedding model configured (see Configure AI features). Without one, search uses keyword matching only.

Where search appears

Help center homepage: the main search box.
Widget Help tab: search inside your app, without leaving it.
Admin article list: filter your own articles while managing content.

Search only returns published articles the visitor is allowed to see, based on category visibility and audience segments.

Ask AI

When AI is available, the search box also offers Ask AI. A visitor types a question and gets a short answer written from your published articles, with the source articles listed underneath. If your articles don't cover the question, Ask AI says so and suggests browsing the articles instead of guessing.

How embeddings are generated

When you create or update an article, Quackback generates its embedding in the background from the title and content. There is nothing to trigger by hand and nothing separate to configure for the help center.

See what visitors search for

In Admin > Help Center, open Article performance (under Other in the sidebar). The search terms table shows each query, how often it was searched, how many searches returned no results, and when it was last searched. Queries with no results are a ready-made list of articles to write.

Write articles that get found

Use clear titles. Titles carry the most weight. "Reset your password" beats "Account access issues" for someone searching "forgot password".
Write the description. It appears in results and helps visitors pick the right article.
Use headings. They structure long articles and add searchable text.
Phrase titles the way customers ask. If they'd type "how do I cancel my subscription", a title like "Cancel your subscription" matches directly.