Copilot Studio: Does semantic grounding inherit SharePoint re-indexing behavior from Microsoft 365 Copilot?

Question

Copilot Studio: Does semantic grounding inherit SharePoint re-indexing behavior from Microsoft 365 Copilot?

Mats 20

I am working for a Swedish municipality, piloting and evaluating Copilot Studio for a number of Q&A-style Agent use cases. The knowledge source document in SharePoint are frequently revised, and automatic re-indexing is a hard requirement. I have tested the following approaches:

SharePoint URL–based knowledge sources → fast to connect, but retrieval quality is shallow and largely lexical.
Dataverse-based upload / “upload and sync” from SharePoint → improved ingestion, but semantic retrieval is still unreliable for indirect questions (e.g. location/contextual queries).
Overwriting a SharePoint file does not seem to consistently trigger re-indexing.

This suggests that:

Dataverse-ingested files are fully reprocessed on metadata and content changes, including automated re-indexing.
SharePoint-based sources are not.

Question: Microsoft documentation for Microsoft 365 Copilot semantic indexing states that updated SharePoint documents are automatically re-indexed.When Copilot Studio semantic grounding is enabled (which requires Microsoft 365 Copilot licensing), is this also intended to rely on the same semantic indexing pipeline and re-indexing mechanism used by Microsoft 365 Copilot, and therefore automatically re-index updated SharePoint documents?

I am not asking about latency guarantees, only whether this behavior is at all intended and supported. Any clarification or documentation reference would be appreciated.

Mats 20 Ryktespoäng

2026-01-12T11:21:22.1133333+00:00

Update: After delayed processing, SharePoint overwrite did eventually trigger re-indexing via Dataverse sync. The behavior is asynchronous and not observable in the UI. The question remains whether Copilot Studio semantic grounding is intended to rely on the same semantic indexing pipeline and re-indexing mechanism as Microsoft 365 Copilot.

Svar godkänt av frågeställaren

0 ytterligare svar

Ditt svar

Mats 20 Ryktespoäng

2026-01-12T11:21:22.1133333+00:00

Update: After delayed processing, SharePoint overwrite did eventually trigger re-indexing via Dataverse sync. The behavior is asynchronous and not observable in the UI. The question remains whether Copilot Studio semantic grounding is intended to rely on the same semantic indexing pipeline and re-indexing mechanism as Microsoft 365 Copilot.

Answer 1

Huy-K 13,990 Externa Microsoft-medarbetare Moderator

Note: As the thread is posted on a Swedish link but written in English, I will kindly respond in English for consistency.

Dear @Mats,

Thank you for posting your question in the Microsoft Q&A forum.

Based on your description:

When Copilot Studio semantic grounding is enabled (which requires Microsoft 365 Copilot licensing), it relies on the same Microsoft 365 Search/semantic indexing pipeline used by Microsoft 365 Copilot. As a result, updated SharePoint Online documents are automatically re‑indexed via that pipeline; Copilot Studio does not provide or require a separate manual re‑index action. Microsoft field guidance also notes that changes to already‑indexed documents are re‑indexed immediately, while new site‑inherited content is indexed on a scheduled basis. Copilot Studio does not expose indexing status in the UI.

Here are some references:

If the answer is helpful, please click "Accept Answer" and kindly upvote it. If you have extra questions about this answer, please click "Comment".

Note: Please follow the steps in our documentation to enable e-mail notifications if you want to receive the related email notification for this thread.

Mats 20 Ryktespoäng

2026-01-13T09:55:12.5266667+00:00

Thanks a lot for your clarification. Sounds very promising indeed. I have a follow-up question: Should we, using semantic grounding for Sharepoint files, expect the same response quality as with files uploaded to Dataverse directly from a drive?
Cheers/Mats
Mats 20 Ryktespoäng

2026-01-13T13:48:18.2666667+00:00

Update: After some more tests, I discovered that the response quality actually increased gradually some time after creating the topic and connecting it to the (previously uploaded and indexed) Sharepoint-based knowledge source. The time lag is not a big problem, as long as we are aware of it. However, I am now pondering if we actually need the Tenant graph grounding (requiring M365 Copilot licensing) at all when uploading the Sharepoint files to Dataverse? Perhaps the Tenant Graph grounding is only needed to improve the search results (via semantic search) when connecting (i.e. not uploading) Sharepoint sources via a Sharepoint Library URL? Cheers/Mats
Huy-K 13,990 Ryktespoäng Externa Microsoft-medarbetare Moderator

2026-01-14T01:38:48.8+00:00
Dear @Mats,
Here is the information you need:

SharePoint URL and Tenant graph grounding (requires an M365 Copilot license and a configured semantic index) uses the Microsoft 365 Search/semantic index pipeline and typically yields higher‑quality, security‑trimmed answers.

Dataverse uploads/sync use Dataverse search/embeddings and do not require Tenant graph grounding; quality is driven by document structure and metadata (e.g., synonyms/glossary).

Indexing freshness may lead to small, visible improvements after connecting SharePoint sources; Copilot Studio does not show indexing status, and Microsoft does not provide latency guarantees.

Guidance: Keep Tenant graph grounding ON for SharePoint URL sources; use Dataverse uploads for curated sets.