Answered

Copilot Studio Agent Finds SharePoint Documents but Fails to Use Content in Responses

(2) Share

Report

Posted on by JN-11120230-0

Issue Summary:

Copilot Studio agent cannot extract content from specific SharePoint documents despite successfully locating them during knowledge retrieval. The agent lists the documents under "Other sources searched over" but returns responses using only the base LLM's general knowledge, indicating content extraction or semantic matching failure.

Environment:

Platform: Microsoft Copilot Studio agent with SharePoint knowledge sources
Document type: .docx files in SharePoint document library
User permissions: Owner membership on the SharePoint site
Knowledge source: Explicitly configured in Copilot Studio

Observed Behaviour:

Documents from the problematic SharePoint site are successfully indexed and discovered by the agent in the Knowledge tab

When queried specifically to target the documents, the agent identifies documents in the SharePoint and lists them under "Other sources searched over"

The agent fails to extract usable content from these documents

Response defaults to LLM general knowledge with message: "This response was generated by using general knowledge from the LLM"

Key Evidence:

Working scenario: Identical documents copied to a different SharePoint location work correctly—agent successfully retrieves and references their content

M365 Copilot comparison: The same documents are successfully surfaced as knowledge sources in standard M365 Copilot chat, proving they are searchable and accessible

Consistent pattern: Issue affects all documents in the specific SharePoint site/library, not isolated files

Troubleshooting Already Completed:

Knowledge source correctly configured in Copilot Studio
Tested direct queries against known document content
Verified document format compatibility (.docx, .pdf)
Confirmed issue is location-specific (same files work elsewhere)
Re-added knowledge source
Used direct file/library paths rather than site root
Recreated the agent from scratch
Elevated SharePoint membership to make user account an Owner
Confirmed that SharePoint site/library indexing is enabled and not excluded from search
Confirmed that account has access at a folder and file level

Requested Support:

Please can someone explain why document discovery succeeds but content extraction fails for this SharePoint location and how to fix it?

Categories:

General topics

I have the same question (3)

All responses (6)

Answers (1)

Romain The Low-Code... 2,315 Super User 2025 Season 2 on at

Like (0)

Report
Copy link

Link copied!

hello,

this other way if undexing use this system : https://learn.microsoft.com/en-us/microsoft-copilot-studio/knowledge-unstructured-data

it's totally hidden but it's like this :

the main difference with standard RAG is the capacity to indexe with better result more complexe data like excel files or unstructured pptx and image with texte.

But Access limitation are the same.

So without more info it's complicated to tell you the difference.

Only one time i have seen some block with ip restriction few years ago.

Just be carefull and read the link below, this way of indexation have few consequences and drawback , few sample :)

- number of files per knolwedge source (if more than 1K files you have to add many KB in SP to bypass it)

- dataverse as index will take a little "size" in the technical environnement in power platform, so could consume space :)

but this way give you better resultt

1 people found this reply helpful.

Was this reply helpful? Yes No
JN-11120230-0 9 on at

Like (0)

Report
Copy link

Link copied!

Hi Romain,

Thank you so much for your response.

See my answers to your questions / comments below in green:

- only one sharepoint lib have problem - correct

- you can reproduce in a fresh blank agent (because some time old agent could have problem with all new feature, very very rare problem) - correct

- User have real access to the library - yes

- security label on the library : do you have any security label - no security labels

PS : did you try with the other sharepoint indexing capacity : not the big button in knowledge, but the small one above witch will use a partial ingest in dataverse - this worked!!!

I am very happy that it is working, but do you know why this worked? What can it tell me about that SharePoint site?

Many thanks again!

Was this reply helpful? Yes No
Verified answer

Romain The Low-Code... 2,315 Super User 2025 Season 2 on at

Like (0)

Report
Copy link

Link copied!

hello

To my opinon and if understand well :

- only one sharepoint lib have problem

- you can reproduce in a fresh blank agent (because some time old agent could have problem with all new feature, very very rare problem)

So the problem is the library

What to check :

- User have real access to the library

- security label on the library : do you have any security label

PS : did you try with the other sharepoint indexing capacity : not the big button in knowledge, but the small one above witch will use a partial ingest in dataverse

2 people found this reply helpful.

Was this reply helpful? Yes No
JN-11120230-0 9 on at

Like (0)

Report
Copy link

Link copied!

Thank you for your response, however this does not answer my question. As mentioned above, the files I am trying to access are word documents (.docx).

Was this reply helpful? Yes No
Nivedipa-MSFT on at

Like (0)

Report
Copy link

Link copied!
Generative answers that rely on SharePoint can only process a limited range of file types. Specifically, supported formats include:

SharePoint pages (note: .aspx pages are not supported)

Word documents (.docx)

PowerPoint presentations (.pptx)

PDF files (.pdf)

For more details, refer to the documentation: Generative answers with Search and summarize - Microsoft Copilot Studio | Microsoft Learn

If you need to include Excel documents in your generative answer node, you can upload the file directly into your Copilot, which accepts a broader variety of file formats, such as:

Word (.doc, .docx)

Excel (.xls, .xlsx)

PowerPoint (.ppt, .pptx)

PDF (.pdf)

Text files (.txt, .md, .log)

HTML (.html, .htm)

CSV (.csv)

XML (.xml)

OpenDocument formats (.odt, .ods, .odp)

EPUB (.epub)

Rich Text Format (.rtf)

Apple iWork files (.pages, .key, .numbers)

JSON (.json)

YAML (.yml, .yaml)

LaTeX (.tex)

You can find the full list of supported document types in the following resource: https://learn.microsoft.com/en-us/microsoft-copilot-studio/nlu-documents#supported-document-types

Was this reply helpful? Yes No
Nivedipa-MSFT on at

Like (1)

Report
Copy link

Link copied!

Thanks for bringing this issue to our attention.

Was this reply helpful? Yes No