Are images and child pages indexed when a URL enters a Mistral Library?

The Citead team, updated

In short. A Mistral Library indexes the text of the supplied webpage only, excluding child pages, images, and charts.

Definition: A Mistral Library is a persistent knowledge base built from documents and web pages. When a webpage URL is added, Mistral indexes the text content of that page, while child pages, images, and charts are excluded.

Key facts

  • Mistral describes Libraries as persistent knowledge bases built from documents and web pages in its Libraries documentation.
  • The webpage workflow is to select Webpage, enter one or more URLs, and select Index.
  • The indexed unit is the text content of the supplied page.
  • Child pages, images, and charts are excluded from webpage indexing.
  • Mistral's separate Open URL documentation says that Open URL fetches only the single supplied page and does not crawl the entire site or follow links.

What enters a Mistral Library?

The Libraries workflow accepts an external webpage as a Library item. Mistral documents three interface actions: select Webpage in the Library view, enter one or more URLs, and select Index. The page then becomes part of a persistent knowledge base that Work can search when the Library is attached to a task. Mistral describes this search as operating across all documents in the Library and returning answers based on the stored data with direct references to source material.

The URL is therefore an input for one webpage item. Mistral documents the Library as a collection searched across its documents when attached to a task.

Are images and child pages indexed?

No. Mistral's Libraries documentation states that only the text content of the page is indexed. Its exclusion sentence names child pages, images, and charts. A link from the supplied page to another page therefore does not add that child page to the Library, and visual material in the supplied page is outside the documented webpage indexing scope.

This is a content boundary for webpage ingestion. It is separate from the fact that a Library can also contain uploaded files. The documentation says that Libraries accept the same formats as file uploads and lists PNG, JPEG, WebP, and GIF under Images.

What can a site owner do when visuals matter?

Mistral gives a documented alternative for full visual content: download the page and upload it as a file instead. That path treats the downloaded file as an upload, rather than relying on the webpage URL's text-only indexing behavior. The supported formats section lists common document, image, text, data, code, and email formats, including PNG, JPEG, WebP, GIF, PDF, Markdown, JSON, and XML.

How is this different from Mistral Open URL?

Open URL is a separate Work feature for a page whose link is already available. Mistral's Search the web documentation says that Open URL lets Work fetch, read, and use the page content as context. Its limitations section specifies that Open URL fetches only the single page at the supplied URL and does not crawl the entire site or follow links.

That wording describes Open URL's retrieval behavior, while the Libraries page describes Library ingestion. For the question about a URL entering a Library, the direct rule is the Libraries page's text-only scope and its explicit exclusion of child pages, images, and charts.

FAQ

Can you add more than one webpage URL to a Library?

Yes. The Libraries instructions say to enter one or more URLs before selecting Index.

Will a chart embedded in the page be available to the Library?

No. Mistral names charts among the content excluded from webpage indexing and directs readers who need full visual content to download the page and upload it as a file.

Should you use Open URL to make a Library crawl a site?

No. Open URL is a separate feature, and its documentation limits fetching to the single page supplied and says it does not follow links.

Sources

  1. Mistral, "Add context with Libraries", date not shown
  2. Mistral, "Search the web", date not shown

Check how a page reads to answer engines. Open the GEO score analyzer

Test a page and see what to fix first

The analysis usually takes less than 10 seconds and needs no account.

400,000+ pages scored in the public benchmark