Embedded images inside uploaded documents are not being extracted.

Lianne Datar 0 Reputation points
2026-10-04T05:12:24.97+00:00

When I upload a document (for example, a .docx or PDF) that contains embedded images, the extraction tool reports the file as “ready,” but the extracted content is completely empty. No images appear, and no image metadata is returned. Only direct image uploads (PNG/JPG) work correctly.

What I’ve tested:

Multiple .docx files with embedded images

Multiple PDFs with embedded images

Files that previously triggered the issue

A fresh test today (04 Oct 2026)

In every case, the extraction tool returns an empty result. Direct image uploads still work normally — the issue only affects images inside documents.

Expected behaviour: Embedded images should be extracted or at least appear as image blocks or metadata in the returned content.

Actual behaviour: The tool returns an empty string for the document content, with no images extracted.

Status: This issue was reported weeks ago, but the behaviour has not changed. The extraction tool still fails to extract embedded images.

Request: Could you please confirm whether this fix is still in progress, and whether there is an updated timeline for resolving the embedded‑image extraction issue?

Microsoft Copilot | Other
0 comments

2 answers

Sort by: Oldest
Most helpful Newest Oldest
  1. Kai-H 27,975 Reputation points Microsoft External Staff Moderator
    2026-10-04T10:04:49.8133333+00:00

    Hi, Lianne Datar

    Since direct PNG and JPG uploads work, I wouldn’t ask you to repeat those tests.

    I can’t confirm from the information available here whether a fix for embedded-image extraction in personal Copilot is still in progress, and I don’t have a verified resolution date. The file-upload guidance lists DOCX, PDF, PNG, and JPEG as supported formats, but it does not promise that images inside a document will be returned as separate image blocks or metadata. So I don’t want to call this an expected limitation or promise a fix without confirmation.

    For now, uploading the images separately is the practical workaround you’ve already verified. To get this investigated, please use the feedback option on the affected Copilot response and include a small sample document, the date of your latest test, and the contrast you found: the document shows as “ready” but returns empty content, while the same image works when uploaded directly. If the document contains private information, use a new sample made only for testing. Feedback with specific examples helps the product team investigate the behavior. I’m sorry I can’t give you the timeline you asked for. I’d rather be clear about that than offer a date that hasn’t been confirmed.

    Please be advised that this is a user-to-user support forum. Moderators and contributors, including external Microsoft employees, are not allowed to directly interfere with Microsoft product features or access back-end systems. Our role is limited solely to providing technical guidance on reported incidents, requests, or ideas.  

    Was this answer helpful?

  2. Lianne Datar 0 Reputation points
    2026-10-05T08:06:26.79+00:00

    Unfortunately for my work I may need to send up to 40 images for one analysis. This is impossible with a 20 file per day limit.

    I did end up on an exhausting 2 hour chat with support who finally escalated to the next level. I have previously left feedback many times and been in a support chat 2 weeks ago with the promise of help. It seems the issue has been largely ignored. Since the last chat was grueling and involved 8 documents sent, I am hopeful that the issue will at least be looked into.

    I appreciate you replying and offering assistance.

    Was this answer helpful?

    0 comments

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.