Does XCC (SharePoint) index full document content, or just metadata for search-then-fetch?

[JK9903]

Hi all,

Question on how the External Content Connector (XCC) for SharePoint actually stores crawled data:

1. Once XCC crawls and indexes a document, is only the metadata (title, URL, permissions, etc.) stored in the index, or is the actual document content also indexed?

2. If it's only metadata — do I need to make a separate API call to SharePoint using that metadata to retrieve the actual document content?

3. If content is indexed, how can I confirm/verify this, and can that indexed content be pulled and used directly within a Now Assist Virtual Agent (NAVA) conversation or an AI agent — or is XCC's indexed data only usable through the standard AI Search experience?

End goal: I want to use this feature so that NAVA / a custom AI agent can read and use SharePoint document content directly, not just surface it as a search result link.

Any insight from folks who've dug into this at the connector/indexing level would help a lot.

[Vikram Reddy · Accepted solution]

Hello @JK9903,

XCC does index actual document content, not just metadata. ServiceNow's own connector description for the SharePoint XCC states plainly that the connector "can search contents of the files on SharePoint," and that only works if the crawler pulls the document text into the AI Search index alongside title, URL, and permission mapping. It is not a search-then-fetch model where you get a pointer and have to go back to Graph API for the body.

The clearest proof is Now Assist in AI Search itself: the Q&A Genius Result card runs a real RAG pipeline, retrieval, then augmentation, then generation, and the augmentation step feeds actual document content into the LLM. That would not be possible if only metadata were indexed.

To verify and use this in practice:

• Open the SharePoint Indexed source and check which fields are marked as semantically indexed, it should be more than just title and URL.
• Run a Q&A query in Global Search or Virtual Agent against a SharePoint doc and see whether the answer summarizes actual content rather than just linking out.
• For NAVA, this same Genius Result mechanism already surfaces summarized content, not just a link.
• For a custom AI agent, add the AI Search retrieval tool in AI Agent Studio, point it at your search profile and the SharePoint source, and select the fields you want returned, that pulls indexed content in-process with no external call needed.

One caveat: the indexed text lives in the AI Search index itself, not in a queryable GlideRecord table, so there is no supported way to GlideRecord-query the raw crawled body outside these AI Search consumption paths.

References

• External Content Connectors Sharepoint Online - ServiceNow Store
• AI Search Retrieval Augmented Generation (RAG)
• Under the hood: Now Assist in AI Search
• AI Agent tools - Getting the most out of your agentic workflows
• External Content Connector - SharePoint - Where are index documents stored?

Thank you,
Vikram Karety
Octigo Solutions INC