PowerLeap Logo

Copilot Studio Friction

Get notified when this record changes

One email when the status or the fix changes — double opt-in, no tracking, unsubscribe in every email.

Your PDF files are ignored or not answered

Opensince 8 July 2026

Last verified

Details & related

Assessment

Confidence
Confirmed officialConfirmed by Microsoft documentation or an official statement.
Severity
Degrading
Typical time lost
Hours

Identification

Knowledge & groundingLicensing & costAll channels

Verification & changes

  1. Verified

    Doc check by human: Official troubleshooting doc re-read during seed migration; 7-MB/200-MB caps, Enhanced-search coupling and top-3 limit still documented.

  2. Change

    Provisionally approved by the Product Owner; external LLM quality review pending.

  3. Change

    Initial record created from the seed dossier (migration wave 1).

Related

Are you in the right place?

  • Your agent answers from some documents, but questions about certain PDFs come back empty or generic.
  • The affected files were added to the knowledge source without any error.
  • Branch by size: the ignored files are noticeably larger than the ones that work.
  • Branch by type: the ignored files are scanned, encrypted, or carry a sensitivity label.
  • Files of other types from the same source are answered normally.

If instead no source returns anything at all → see Knowledge returns nothing — and no error is shown. If instead the SharePoint source won't even save → see Can't add SharePoint source. If instead the missing file was added only recently → see New or updated files don't show up in answers. If instead the agent answers confidently but wrongly → see Agent makes up answers or uses outdated information.

What's happening

Cause 1: a licensing file-size cap. Your tenant's license decides how big a file may be. Without a Microsoft 365 Copilot license in the tenant, the cap is 7 MB. With that license, the cap rises to 200 MB. Both figures are documented as of 2026-07-08. The "Enhanced search results" setting must match: off without the license, on with it. Generative answers find bigger files in search but never read them. Picture a mailroom scale: parcels over the weight limit are returned unopened.

Cause 2: the file is not machine-readable. Scanned paper PDFs contain pictures of text, not text. Encrypted files, IRM-protected files, and files with a sensitivity label are skipped the same way. IRM means Information Rights Management, a Microsoft protection feature. The agent cannot read their content, so it treats them as absent. No warning marks the skipped files.

For technicians

Official troubleshooting, verbatim: "if you don't have a Microsoft 365 Copilot license in the same tenant as your agent, generative answers can only process files up to 7 MB in size. You must also turn off the Enhanced search results feature."

  • With the license, the documented maximum is 200 MB, with "Enhanced search results" turned on (documented as of 2026-07-08).
  • Only the top three SharePoint search results are summarized, documented as of 2026-07-08.
  • Both causes fail without any error surfaced to the maker or the user.

How to fix it

Solution 1

Official workaround

Work within the 7-MB cap (no Microsoft 365 Copilot license)

  1. Confirm with your admin that your tenant has no Microsoft 365 Copilot license.
  2. Turn the "Enhanced search results" feature off for your agent.
  3. Identify every knowledge file larger than 7 MB.
  4. Split each oversized PDF into parts smaller than 7 MB.
  5. Replace the oversized originals in the source with the parts.

✅ You should now see: questions about the split files answered in the test pane.

Solution 2

Official workaround

Use the 200-MB cap (with a Microsoft 365 Copilot license)

  1. Confirm the Microsoft 365 Copilot license exists in the same tenant as the agent.
  2. Turn the "Enhanced search results" feature on.
  3. Keep individual files below 200 MB.

✅ You should now see: previously ignored large files being used in answers.

Solution 3

Community workaround

Make protected or scanned files readable

  1. List the files that are scanned, encrypted, IRM-protected, or labelled.
  2. Recreate scanned files as text PDFs with OCR software. OCR (optical character recognition) turns pictures of text into real text.
  3. Ask your admin whether sensitivity labels on knowledge files can be adjusted.
  4. Replace protected originals with readable copies where your policies allow it.

✅ You should now see: questions about the reworked files returning grounded answers.

Check that it worked

Pick a question whose answer only exists in a previously ignored PDF. Ask it in the test pane. Expected: a grounded answer that cites the file.

If it didn't work

  • The file still exceeds your cap. Check the exact file size; the cap counts the file, not the text.
  • Enhanced search doesn't match your license. The setting must be off without the license and on with it.
  • Your answer sits outside the top results. Only the top three SharePoint results are summarized. Consolidate content into fewer, well-named documents.
  • The change hasn't reached your channel. Publishing propagates slowly. See Teams answers with an old version of your agent.

Prevent it next time

  • Run a document audit before you onboard a library: file sizes, scans, labels, protection.
  • Record the tenant's license situation in your project notes before you plan knowledge features.
  • Treat the license question as a planning gate, not a debugging discovery.

Evidence

  • Official documentationlearn.microsoft.com

    States the 7-MB cap without an M365 Copilot license, the 200-MB cap with one, the Enhanced-search coupling, and the top-three-results summarization limit.

  • Official documentationlearn.microsoft.com

    Documents SharePoint file handling for generative answers, including that content must be machine-readable to be used.

  • Official documentationlearn.microsoft.com

    The quotas page lists knowledge source limits, including file-size caps, as of the retrieval date.