Share via

Copilot Notebook reference size limit

Scian, Paul 40 Reputation points
2026-03-03T21:17:52.3833333+00:00

What is the current size limit of a reference added to a Notebook in copilot

Microsoft Copilot | Microsoft 365 Copilot | Development
0 comments No comments

1 answer

Sort by: Most helpful
  1. AI answer

    2026-03-03T21:18:02.6866667+00:00

    For Copilot Notebooks, the documented limits apply to the number and type of references, not to a specific per-file size cap unique to Notebooks:

    • Up to the first 100 references in a notebook are used for grounding; additional references beyond the first 100 are ignored for grounding.
    • Supported reference file types are: .docx, .pptx, .xlsx, .pdf, .loop, .page, and OneNote pages.

    For file size and length, Copilot behavior follows the general document-length guidance used across Microsoft 365 Copilot rather than a Notebook-specific size limit. Relevant guidance includes:

    • When summarizing or referencing content, Copilot works effectively up to about 1.5 million words or 300 pages.
    • For Q&A over a document, results work best if the document is under ~7,500 words.
    • For rewrite scenarios, documents work best if under ~3,000 words.

    In extensibility scenarios where SharePoint files are referenced (for example via agents), additional practical limits apply:

    • For SharePoint files referenced by URL, keep files to a maximum of 36,000 characters (~15–20 pages) to help Copilot reliably retrieve the right content.
    • For embedded file content, Copilot indexes the first 750–1,000 pages (~1.8 million characters) of each embedded file.

    These limits describe how much of a referenced file Copilot can effectively read and use, rather than a hard “upload size” limit specific to Notebook references.


    References:

    AI-generated content may be incorrect. Read our transparency notes for more information.

    Was this answer helpful?

    1 person found this answer helpful.

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.