Optical character recognition across the Microsoft ecosystem now spans several applications and services, from SharePoint to Purview and Copilot. Each component serves a different purpose, with its own activation requirements and limitations. Microsoft OCR in Office and SharePoint is not a single feature: it is a set of mechanisms that work in parallel, sometimes without users even realizing it.
Search OCR and Purview compliance OCR: two distinct mechanisms
A common misconception is to treat SharePoint OCR as a single, all-in-one feature. In reality, two separate processes coexist. The first extracts text from images and scanned documents so they can be found through SharePoint search. The second, in Microsoft Purview, analyzes images to enforce compliance rules, including data loss prevention (DLP).
Administrators must enable each process separately. Purview OCR is optional and requires dedicated configuration: you need to choose which locations to scan and configure the associated DLP policies. Enabling OCR for search does not mean images will be analyzed for compliance purposes.
To learn more about how Microsoft OCR in Office and SharePoint works, it is worth paying particular attention to the licensing differences between these two use cases, as they determine what is actually available with each subscription.

SharePoint indexing: what OCR does not guarantee
Uploading a scanned PDF to a SharePoint library with OCR enabled does not make the document immediately searchable. Indexing takes place in the background, with no guaranteed turnaround time, depending on service load and content volume. There is no setting that forces instant processing.
This delay has practical consequences for organizations handling large volumes of scanned documents. An accounting department that uploads fifty scanned invoices may find that some do not appear in search results for several hours.
Scan quality and extraction accuracy
Microsoft notes that character and scan quality directly affect OCR results. A document photographed at an angle, with uneven lighting or low resolution, will produce incomplete or inaccurate extracted text. Handwritten forms pose a similar challenge: OCR can process handwriting, but recognition rates drop when the writing is difficult to read.
Three factors can reduce extraction quality:
- Low contrast between text and background, common on preprinted forms or colored industrial labels
- Document rotation or distortion, even slight, which makes it harder for the OCR engine to segment text into lines and words
Copilot and OCR: an additional layer to understand
The arrival of Copilot in Microsoft 365 adds another dimension to processing scanned documents. Copilot can summarize PDFs, including those processed with OCR, and answer questions about their content. This capability relies on the extracted text layer: without prior OCR, Copilot cannot read a purely image-based PDF.
The distinction between a native PDF (created from a word processor) and a scanned PDF (created with a scanner) therefore remains crucial. A native PDF already contains a text layer. A scanned PDF that has not undergone OCR remains opaque to Copilot and SharePoint search alike.
Known limitations of Copilot processing
Extraction accuracy varies by language and document type. Multilingual documents, or those combining printed text with handwritten annotations, remain edge cases where the results may need to be checked manually.
Copilot works in OneDrive, SharePoint, Teams, and Edge, but the underlying OCR depends on whether it has been enabled at the tenant level. If OCR is not configured in the SharePoint admin center, Copilot will not have the text layer it needs to process scanned documents.

Classifying historical content: a frequently overlooked step
Organizations adopting SharePoint OCR tend to focus on incoming documents. Existing archives, sometimes containing thousands of files already stored in libraries, are not automatically reprocessed. On-demand classification in Purview can help catch up, but it requires the administrator to take explicit action.
This point is rarely covered in getting-started guides. A company that enables OCR today makes future documents searchable, but its archives remain unchanged until reprocessing is initiated. The effort required depends on the volume, format, and organization of the existing files.
Formats supported by SharePoint OCR
Files that already have a text layer are not reprocessed. This selective approach means OCR does not modify native PDFs or create duplicate text layers.
OCR in the Microsoft ecosystem relies on several components that must be enabled, configured, and monitored separately. There is no single button that makes every document searchable at once. The distinction between search OCR, compliance OCR, and Copilot processing determines what users can actually expect from the system.



