About

In Content Hubs, you can search across your whole site with global search, or narrow things down by searching within a specific area, like My Media, a channel, or a gallery. The search options and filters you see depend on where you're searching.

This article explains how Content Hubs decides which results appear first, and how your metadata (like titles, descriptions, and tags) and search settings affect that order.

At a glance

• What matters most: Media name, description, tags, reference ID, and custom metadata have the greatest influence on relevance. See Metadata tips.
• What affects result order: Field weights, how and where a term matches, and other relevance signals contribute to ranking. If a result isn't where you expect it to be, see Why did I get these results?
• What doesn't affect relevance: Views, plays, likes, and comments don't contribute to the relevance score.
• What administrators can configure: Which sources are searchable and which filters users see. See For administrators.

What content is searchable?

Content Hubs uses eSearch, Kaltura's search service, to find and rank matching content. Depending on the search context and your site configuration, eSearch can search across different types of metadata and content, including:

• Media name
• Description
• Tags
• Reference ID
• Custom metadata
• Owner and collaborator information
• Chapters and slides
• Q&A, polls, and annotations
• Captions
• OCR data (text detected in video frames)

Administrators can configure some of these searchable sources. For more information, see Search module and Customdata module.

How search results are ranked

You don't need to know the details to get good results, but here's what happens behind the scenes.

eSearch calculates relevance by combining Elasticsearch's BM25 algorithm with Kaltura-specific weights and boosts.

How BM25 scores matches

BM25 evaluates three primary signals across your site's content:

• Term frequency: How often the search term appears in a media entry or field.
• Term rarity: Terms that are rare across your site's content carry more weight than common terms.
• Field length: A match in a shorter field can be more significant than the same match in a longer field.

How Kaltura applies field weights

Kaltura also assigns a weight to each searchable field and applies it to the BM25 score for matches in that field. In general, a match in a higher-weight field has more influence on ranking than the same match in a lower-weight field:

Searchable content	Weight	Relative influence
Media name, reference ID, tags, description, custom metadata	100	High
Owner and collaborator information, chapters and slides, Q&A, polls, and annotations	50	Medium
Captions	10	Low

These values (100, 50, and 10) are field weights used in the relevance calculation, not final scores assigned to an item. The Search module doesn't provide settings for modifying these baseline weights.

Media name, reference ID, tags, description, and custom metadata have the highest relevance weight. When defining metadata standards, prioritize clear and consistent information in these fields.

How text matches affect relevance

eSearch evaluates multiple types of text matches and combines them when calculating relevance. Exact matches receive more weight than less direct matches. eSearch also accounts for different languages, synonyms, and partial matches.

The number of times a search term appears also contributes to relevance, but repetition alone doesn't determine ranking. Several matching signals contribute to the final relevance score.

Why did I get these results?

If search results aren't what you expect, the following factors can help explain why:

What you see	Why it can happen
An item appears even though the term isn't in its name.	The term matched other searchable content, such as its description, tags, custom metadata, captions, chapters, or OCR data.
Two items containing the same term rank differently.	Where and how the term matches affect relevance. For example, matches in the media name have a higher field weight than matches in captions. BM25 also considers factors such as field length and how common the search term is across your site's content.
The same search returns a different result order than before.	Changes to the content library can affect how common a term is across your site's content, which can change BM25 scores and result order.
You see different results when you search in different areas.	The content and filters available can vary depending on where you search.

Relevance and other sort options

Relevance puts the best matches first.

Depending on the search context, users can also sort results by creation date, last updated date, alphabetical order, views, plays, likes, and comments.

Available sort options can vary depending on the search context and site configuration.

Metadata tips

Use clear and consistent metadata to help users find the right content.

Metadata	Do	Avoid
Media name	Use a clear, specific name that describes the subject and includes terms users are likely to search for (for example, 2026 Product onboarding for administrators).	Generic names that provide little context, such as Training session 4.
Description	Add useful context and terminology that isn't clear from the media name.	Repeating keywords unnaturally only to influence ranking.
Tags	Use consistent terms for meaningful topics, product names, and other information users are likely to search for.	Creating multiple variations or spelling variants of the same tag.
Custom metadata	Use consistent, structured values across media for important schema fields.	Leaving important fields blank or inconsistently populated.
Captions	Use accurate captions to make spoken content searchable when caption search is enabled.	Relying on captions alone instead of providing an accurate media name, description, tags, and custom metadata.
Keywords	Use accurate, natural terms within the description and tags.	Repeating the same keywords over and over. Repetition alone doesn't determine ranking.

For administrators

In the Search module, you can configure whether tags, media owners, reference IDs, captions, and OCR data are included in search.

You can also configure which filters are available to users.

Custom metadata search is configured in the Customdata module. Other search options, such as Kwebcast fields and ownership filters, are configured in their respective modules.