# SharePoint OCR: Can It Search Your Scanned Documents?

- Author: Schema Law (https://www.schemalaw.com/)
- Published: 2026-10-09T18:43:13Z
- Updated: 2026-10-11T19:22:20Z
- Category: OCR in everyday software
- Canonical: https://www.schemalaw.com/blog/sharepoint-ocr

> SharePoint OCR is a paid Microsoft 365 service, not a default. What it reads, which scanned documents it skips, who turns it on and how it is billed.

SharePoint OCR does not run on scanned documents by default. Microsoft sells optical character recognition for SharePoint as a pay-as-you-go service that an administrator starts by linking an Azure subscription. Once it is on, text in newly uploaded scans and images is indexed for search. Scans that were already in the library are not processed.

That answer has several parts worth checking against your own tenant, because a firm can believe its scans are searchable for years before someone looks for a word and gets nothing. Everything below about what SharePoint does comes from Microsoft's own documentation, linked where it is used. This post covers SharePoint in Microsoft 365. If your firm runs SharePoint Server on its own hardware, the pages cited here do not describe your system.

## Does SharePoint OCR scanned documents on its own?

No. SharePoint stores a scanned PDF and lists it in the library, but it reads the words on the page only if your organization has set up Microsoft's optical character recognition service. Microsoft's setup page calls OCR "a pay-as-you-go service that is set up in the Microsoft 365 admin center". Without that step, a scan is a picture to SharePoint.

Three things follow from Microsoft's page on [setting up and managing optical character recognition](https://learn.microsoft.com/en-us/microsoft-365/documentprocessing/ocr):

- **It needs billing first.** Before anyone can use the service, an administrator has to link an Azure subscription.
- **It then switches itself on widely.** Microsoft says that after the subscription is linked, OCR "will be automatically set up and enabled for all SharePoint sites". Nobody ticks a box per library.
- **It can be narrowed.** An administrator can change the setting from all sites to selected sites or to no sites.

So there are two common states. In a tenant with no pay-as-you-go billing, nothing is recognized. In a tenant where IT linked a subscription for some other document service, OCR may already be running without the legal staff having been told. Ask before you assume either one.

## What Microsoft's optical character recognition service for SharePoint does

Microsoft's OCR service for SharePoint extracts printed or handwritten text from images and documents and puts that text into the search index. It does not hand you a new file. The scan in the library stays as it was, and the recognized words live in SharePoint's index.

Microsoft's [overview of optical character recognition](https://learn.microsoft.com/en-us/microsoft-365/documentprocessing/ocr-overview) gives the sequence in one example: you enable the service, add image files to a document library, and the feature scans them, extracts the text and makes it available for search and indexing. The page adds that the extracted information can also be made available to compliance features such as data loss prevention.

### Where the recognized text goes

The destination depends on the file type, and the difference matters for scans. Microsoft's note on the same page reads:

> When you apply OCR to an image file, the text is stored in the Extracted text metadata column. When you apply OCR to a PDF or TIFF file, the extracted text is indexed in search but not available in the metadata column.

If your scans are PDFs, there is no column where you can read what the service recognized. The only way to see whether a page was read, and how well, is to search for words you know are on it.

### What it does not give you

Microsoft's pages describe indexing. They do not describe writing a text layer into your PDF, and they do not describe exporting the text. A scan that SharePoint has indexed is still an image-only file when you download it, email it or file it. The post on [why you can't copy text from a PDF](https://www.schemalaw.com/blog/why-cant-i-copy-text-from-a-pdf#what-is-an-image-only-pdf) explains what an image-only PDF is and how to recognize one.

### Where it works

Microsoft says the service is available for SharePoint sites, including hub sites, sites associated with a hub site, and the primary site of a site collection, and that subsites are not supported. The overview's table of locations also lists OneDrive accounts and Exchange as places where recognized text is available for end-user search. Teams messages and devices appear in that table for compliance solutions only.

## Which file types and sizes does SharePoint OCR cover?

Microsoft lists PDFs (scanned and hybrid), TIFF, JPEG, PNG, BMP, GIF, HEIC and a long set of camera raw formats for SharePoint and OneDrive. Images must be under 50 MB and between 50 x 50 and 16,000 x 16,000 pixels. Only files uploaded after OCR was enabled are scanned.

The details, all from Microsoft's [requirements and limitations for OCR](https://learn.microsoft.com/en-us/microsoft-365/documentprocessing/ocr-overview#requirements-and-limitations):

| Question | What Microsoft's page says |
| --- | --- |
| Scanned PDFs | Supported, described as ".pdf (scanned and hybrid)" |
| PDFs with some typed pages and some scanned pages | Supported. Newly uploaded hybrid PDFs are processed |
| Scanner image formats | .tif, .tiff, .jpg, .jpeg, .png, .bmp and others |
| Phone photos | .heic and .heif are on the list |
| Pictures inside Word, PowerPoint and Excel files | Extracted and scanned, with a check for duplicate images so the same picture is not charged twice |
| Image size | Less than 50 MB |
| Image dimensions | At least 50 x 50 pixels, no larger than 16,000 x 16,000 |
| Files already in the library | "Images uploaded after OCR has been enabled are the only images that are scanned" |
| Languages | More than 150 |
| Handwriting | The service description says "printed or handwritten text" |

Read three of those rows with care.

**The size and dimension limits are written for images.** The file limitations section does not state a separate size limit for PDFs. Do not assume the 50 MB figure applies to a 300-page production, and do not assume it does not. Test one.

**The backlog row is the one to plan around.** Microsoft's sentence speaks of images uploaded after the service was enabled. The page describes no step for sending existing files through the service. A firm that turns OCR on in October should expect the closed matters scanned in earlier years to stay unsearchable unless something else is done with them.

**Handwriting is listed, not promised.** Microsoft says the service extracts handwritten text. It gives no figure for how well, and this post makes none. Check a page of real handwriting from your own files.

## How SharePoint search treats an image-only PDF

SharePoint search finds a document by the text it can parse from the file and by the file's properties. An image-only PDF contains no characters, so without OCR the search index holds only what is outside the page: the file name and whatever columns someone filled in.

Microsoft's reference list of [file types SharePoint search crawls and parses by default](https://learn.microsoft.com/en-us/sharepoint/technical-reference/default-crawled-file-name-extensions-and-parsed-file-types) includes PDF among the formats with a built-in format handler. A handler reads text that is stored in a file. A typed PDF, or a scan that already carries a text layer, gives it something to read. A bare scan gives it nothing.

### A five-minute test for your own library

1. Pick a scanned PDF that has been in the library for at least a day. Open it and choose a distinctive phrase from the middle of a page, such as a surname next to a date.
2. Confirm the file is image-only. Try to select one word in the PDF viewer. If you cannot, it is a scan with no text layer.
3. Search the library for the phrase in quotation marks.
4. Search for part of the file name.
5. Compare. If the file name finds the document and the phrase does not, the content of that scan is not in the index.

Run the test twice: once on an old scan and once on a scan you upload today. If the new one is found by its content and the old one is not, your tenant has OCR switched on and a backlog that was never read.

### Two search limits that affect long scans

Even with text available, Microsoft's [search limits for SharePoint](https://learn.microsoft.com/en-us/sharepoint/search-limits) cap how much of a large file reaches the index:

- **Download size.** The crawl handles PDFs up to 512 MB. Above that, Microsoft says only the document's metadata is downloaded and the content is not available for search.
- **Parsed content.** Search stops parsing an item after 2 million characters and marks it as partially processed. Content after that point is not indexed.

A consolidated production of several thousand pages can pass the second limit. If a search finds hits early in a long file and none near the end, split the file and upload the parts.

Use the library search box on a distinctive line from the complaint, the label of an exhibit, and a surname from the deposition. When the upload is a discovery set of medical records, search a date taken from the final page of the scan. Record whether each search finds the file, with the PDF open so you can see the words yourself.

## Who turns SharePoint OCR on, and how is it billed?

A SharePoint Administrator or Global Administrator turns it on in the Microsoft 365 admin center, after linking an Azure subscription for pay-as-you-go billing. Microsoft charges per page processed. When this post was written, the published rate was $0.001 per transaction, and a file is processed again each time it is edited.

### Who does what

Microsoft's page on [setting up pay-as-you-go billing for document processing](https://learn.microsoft.com/en-us/microsoft-365/documentprocessing/syntex-azure-billing) lists the prerequisites: an Azure subscription in the same tenant, a resource group in that subscription, an administrator role in Microsoft 365, and owner or contributor rights on the Azure subscription. Whoever holds those roles for your firm, in-house or at an outside IT provider, is the person to ask.

To limit OCR to particular sites, Microsoft's setup page gives this path:

1. In the Microsoft 365 admin center, open Settings, then Org settings.
2. Select Pay-as-you-go services, then the Settings tab.
3. Under Document & image services, select Optical character recognition.
4. Choose Edit, then change All sites to Selected sites (up to 100) or No sites.
5. Save.

### What it costs

The figures below are from Microsoft's [pay-as-you-go pricing for document processing](https://learn.microsoft.com/en-us/microsoft-365/documentprocessing/syntex-pay-as-you-go-services), read on October 8, 2026. Prices change, so open the page before you budget.

- A transaction is one page of an image file, one page of a PDF or TIFF, or one embedded image in a Teams chat or email message.
- The listed price is $0.001 per transaction, in US dollars.
- "Processing occurs every time the file is edited."
- Charges appear on service meters in the linked Azure subscription. Microsoft notes they do not count toward an Azure consumption commitment.

A worked example, using a fictional firm. Hartwell & Osei uploads about 30,000 scanned pages a month to its matter sites. At the listed rate, that is 30,000 transactions, or $30 a month. If staff then edit a quarter of those files and each edit triggers processing again, add 7,500 pages and $7.50. The sums are small. The point of doing them is to know that the meter runs on edits as well as uploads, and on every site unless someone narrowed the setting.

Microsoft also publishes a [trial offer for pay-as-you-go document services](https://learn.microsoft.com/en-us/microsoft-365/documentprocessing/promo-syntex). When read on October 8, 2026, it listed 2,500 pages of OCR a month per tenant at no charge through December 2026, for organizations with pay-as-you-go billing set up. On the same day, Microsoft's OCR overview and its pay-as-you-go pricing page still said the offer ran through June 2026. The pages disagree and this post cannot tell you which is current, so ask your administrator what the Azure meter shows.

To see what you are being charged, Microsoft's billing page points administrators to Cost Management for Azure, filtered by product, and says usage can take up to 24 hours to appear.

### Questions to put to whoever runs your tenant

- Is pay-as-you-go billing linked, and since what date?
- Is OCR set to all sites, selected sites or no sites?
- Do our matter workspaces sit in subsites, which Microsoft says are not supported?
- What have the OCR meter charges been for the last three months?

## Making scans searchable before they reach a SharePoint library

The alternative to recognizing scans inside SharePoint is to upload files that already contain their text. A searchable PDF carries a text layer in the file, which SharePoint's PDF handler can read without the OCR service, and which stays with the document when it leaves the library.

This route answers four gaps in the built-in service as Microsoft documents it:

- **The backlog.** Files uploaded before OCR was enabled are not scanned. Converting the old scans and uploading the searchable versions gets them into the index.
- **The file itself.** Recognition inside SharePoint lives in the index. A text layer lives in the PDF, so the document is still searchable when you send it to a client, attach it to a filing or move it to another system.
- **Seeing what was read.** For PDFs, SharePoint does not expose the recognized text in a column. A text layer can be selected, copied and proofread.
- **Subsites and switched-off sites.** A file with its own text does not depend on where the service is enabled.

It has costs too. Someone has to run the conversion, and you must decide whether the searchable copy replaces the original scan in the library or sits beside it. For records you may have to produce, keep the original as received and name the searchable version so the two cannot be confused.

Many tools can add a text layer, from scanner software to desktop PDF editors, and the guide to [making a PDF searchable without Acrobat](https://www.schemalaw.com/blog/make-pdf-searchable-without-acrobat) compares them. Schema Law is one option: you upload a PDF, PNG or JPEG of up to 20 MB in a browser and download a searchable PDF, described on the [export formats list](https://www.schemalaw.com/#formats) as your original file with a hidden text layer. It does not connect to SharePoint. You download the file and add it to the library yourself, the way you would any other document. For a whole closed-matter archive, the post on [batch OCR for discovery documents](https://www.schemalaw.com/blog/batch-ocr-discovery-documents) covers planning a large run.

One workable division: let SharePoint's service catch the everyday scans that arrive from now on, if your administrator has enabled it and you have tested it, and convert upstream the documents where the file itself has to be searchable. Firms on a legal document management system face the same question, and the companion post on [NetDocuments OCR](https://www.schemalaw.com/blog/netdocuments-ocr) sets out how that platform answers it.

## Try it on a real document

Upload a scanned PDF and export it as Word, a searchable PDF or plain text on the free plan. Your documents are never used to train an AI model. [Start free](https://www.schemalaw.com/api/auth/google)