What changed in 2026
Two items on Microsoft's 365 roadmap matter here:
- Declarative agents can read scanned PDFs. Roadmap item 550514 says newly created declarative agents can ground answers in scanned and image-based PDFs stored in SharePoint. It's listed as launched in August 2026, and it applies to agents created after the change, so an older agent may still miss scans.
- Copilot Chat search is catching up. Roadmap item 559613 says Copilot Chat should better match scanned PDFs and text inside images embedded in Word, Excel and PowerPoint files. In mid-2026 it was still listed as in development.
So the answer depends on where you ask. A new agent pointed at a SharePoint library may read the scan; everyday Copilot Chat may not, yet. Test one known scanned file in your tenant before you rely on either.
Why Copilot skips scanned PDFs
Copilot draws on Microsoft Graph and the search index behind SharePoint and OneDrive. For a document to be answerable, its text has to make it into that index. Born-digital files - a Word export, a PDF from a design tool - index cleanly. Scanned files are pictures, so unless something has run OCR over them, they land in the index with little or no text.
Whether image OCR happens at all in SharePoint depends on your tenant's configuration, and it's inconsistent in practice - which is why the same scanned invoice or signed contract can be searchable in one library and invisible in another.
Quick check: open the PDF from SharePoint and try to select a sentence. If nothing highlights, it's image-only - that's why Copilot has nothing to say about it.
Fix 1 - Make the file searchable
The durable fix is to get real text into the file:
- OCR the PDF and re-upload a searchable copy, so its text lands in the index - see how to make a PDF searchable, or batch OCR for a whole library.
- Ask your SharePoint admin whether image/PDF OCR is enabled for the relevant libraries - this is a tenant-level setting, and turning it on makes newly added scans searchable going forward.
This is the right long-term move for a library that many people search. The downside is that it needs admin involvement and doesn't help you with a file you need an answer from right now.
Fix 2 - Read the file in the browser, with citations
When you just need the answer, you can read the document where it already lives. Sidenote is a browser reading assistant that reads SharePoint and OneDrive files read-only via Microsoft Graph - respecting the access you already have - and OCRs scanned PDFs so you can summarise and question them even when they never made it into the index.
Why it fits this exact problem:
- OCR on open. An image-only scan becomes readable and citable without waiting on a re-index or an admin setting.
- Cited answers. Every claim carries a scroll-to-source citation to the exact sentence, and unsupported claims are dropped server-side - which matters when the "document" is a policy or a contract someone will act on.
- Read-only and permission-aware. It reads the same files you can already open; nothing is written back to SharePoint. See how to summarise a SharePoint document for the full workflow.
Copilot vs an in-browser reader for scanned SharePoint files
| Microsoft 365 Copilot | In-browser reader (Sidenote) | |
|---|---|---|
| Scanned PDFs not yet OCR'd | Invisible - not in the index | OCRs on open |
| Depends on tenant OCR settings | Yes | No |
| Verify an answer | Varies | Click to the exact highlighted passage |
| Works on a single file now | Needs indexing first | Immediately, read-only via Graph |
| Best for | Broad answers across indexed content | Reading and citing a specific file |
The two aren't rivals - Copilot is built for organisation-wide answers across everything that's indexed, while an in-browser reader handles the specific, not-yet-searchable file in front of you. Plenty of teams use both.