Content and indexing

Content shows the documents in the selected Project environment and index with their category: the section label from your import or CMS, or the first folder of a crawled page without one. Search, filter by category and open a document to see how it does in search, which queries show it, how its result looks, its searchable text and its record details.

Add or update content

Use Add content → Import JSON, Crawl website (see Crawler) or Migrate from Algolia (see Migrate from Algolia). Your agent can also prepare content from repository files or a custom source through the Search skill and CLI.

Keep document IDs stable across updates. Provide useful titles, real page URLs and text that contains the information visitors search for. Snippets and section labels improve the result presentation. A source’s administrative name does not automatically become a visitor-facing category.

A full-replacement snapshot replaces the content of the selected named index, not the entire Project. Include every source and document you want to keep in that index. Uploading several replacement files in sequence is not an append operation.

Review source conflicts

A page can occur in more than one source, such as a crawler and a repository import. Review conflicts and choose which source should supply that page before publication. Check the resulting added, changed and removed counts, not only the uploaded file’s size.

Crawler updates retain unrelated sources in the same index. When several sources describe the same page, the chosen source determines its searchable content.

Publish and check the result

Follow the operation in Activity. Accepted input can still be queued or building; the last successfully published index remains available until a new publication succeeds.

Rebuild index rebuilds content already stored in SiteOS. It does not fetch new website pages. To pick up changed source content, run a crawler or prepare a new import first.

After publication, try changed page titles and body phrases in Playground. Review broken destinations and unexpected removals before adjusting ranking.

Curate the empty search

Open Suggestions in the Search menu and use Choose articles to pick and order up to 10 articles from the index. Popular in search, not pinned lists the articles visitors open most from search results in the last 30 days; Pin adds one to your draft. Pinned articles show whether they are among the most opened or rarely opened from search. Content marks the articles that are currently suggested. Preview the selection, then publish it. A connected interface can receive these changes without reindexing content or rebuilding the website.

Suggestions are curated article links shown before a visitor types. They are not query autocomplete or a ranking rule. Removed documents are omitted from the published suggestions; to show no suggestions, explicitly publish an empty selection.

See which articles help

Content can show appearances, clicks and click-through rate for the selected period. Sort or filter the list to find material that appears often but receives few clicks. A document is marked Never shown only once activity covers the whole selected period; right after collection starts, zeros only mean “not yet”. Open an article to inspect its contributing queries. An appearance means the article was returned in recorded results, not that it entered the visitor’s viewport.

Read Search analytics for collection requirements and metric definitions.

Remove a document

Use Remove from index… in a document's menu or at the bottom of its details and review the confirmation. This removes the record from the selected index, not the original website page or another index.

A later source update can restore it. Exclude a crawled page from its crawler, or omit it from the next complete imported snapshot, to keep it out.

These documentation pages and their live Search index are published separately. Deploying an updated article does not automatically import or reindex it.