Import from cloud: Connect Google Drive Connect Dropbox
← All Guides
OCR · Guide · ~5 min read

What is OCR? How It Works.

Scanned PDFs are just pictures of text — you can't search them, copy from them, or convert them to Word. OCR (Optical Character Recognition) fixes that by converting those images into real, usable text.

When someone scans a physical document and saves it as a PDF, the result is essentially a photograph. The scanner captures an image of the page — the text looks like text, but to the computer, it's just a collection of dark pixels on a light background. There's no actual text data, so you can't select words, use Ctrl+F to search, or copy a sentence to paste elsewhere.

OCR solves this. Optical Character Recognition analyses those pixel patterns and identifies which character each cluster represents, then reconstructs a proper text layer in the document. After OCR, your scanned PDF becomes searchable, selectable, and convertible — just like a natively digital document.

How OCR Technology Works

Modern OCR uses neural networks trained on millions of images of text to recognise letter shapes. Here's a simplified version of what happens when you run OCR on a scanned page:

  1. Pre-processing: The scanned image is cleaned up — straightened (deskewed), contrast-enhanced, and noise-reduced — to improve recognition accuracy.
  2. Layout analysis: The OCR engine identifies regions of the page: text blocks, headings, tables, images, columns. This determines how text flows and what order it should be read in.
  3. Character recognition: Each character is analysed individually. The neural network compares each character's pixel pattern against patterns it learned during training and assigns the most likely character.
  4. Language modelling: The recognised characters are assembled into words and sentences. Language models help disambiguate between visually similar characters (like 'l' and '1', or 'O' and '0') based on context.
  5. Output generation: The recognised text is embedded as a new layer in the PDF, either as a visible replacement for the image or as an invisible layer beneath the scanned image (so the document still looks like the original scan, but the text is now searchable).

When Do You Need OCR?

You need OCR whenever you have a PDF that contains scanned pages (photos of paper) rather than digitally created text. The key indicator is: can you select, highlight, or search the text in the PDF? If yes, OCR has already been applied or the document was created digitally. If not, the document is image-based and needs OCR.

Common situations where OCR is needed include documents received by fax or physical post that were scanned in, archival documents digitised from paper records, contracts or forms that were signed on paper and then scanned, invoices from suppliers who print and scan rather than generating PDFs digitally, and any document converted to PDF using a photocopier's "scan to email" function.

What OCR Does and Doesn't Handle Well

OCR handles well:

OCR struggles with:

For better OCR results: Scan documents at 300 DPI or higher (not the lowest-quality scanner setting). Make sure pages are straight in the scanner. Clean scans — good contrast, no shadows — produce dramatically more accurate OCR output than poor-quality scans.

OCR and PDF Conversion

Running OCR before converting a PDF to another format (like Word or Excel) dramatically improves conversion quality. Without OCR, a PDF-to-Word conversion of a scanned document can only convert the page images — you'll get Word files with nothing but pictures on each page, with no editable text. After OCR, the conversion engine has real text to work with and can produce a properly editable document.

If you're planning to convert a scanned PDF to Word, always run it through the OCR tool first, then use the PDF to Word tool for the conversion.

Step-by-Step: How to Run OCR with PDFusion

  1. Go to the OCR tool at fusepdfs.co.uk/ocr.
  2. Upload your scanned PDF — drag and drop or click to browse. Maximum file size is 50 MB.
  3. Click Make PDF Searchable — the OCR process runs on each page. Processing time depends on the number of pages: expect 2–5 seconds per page for typical scans.
  4. Download the OCR'd PDF — the result looks identical to your original scan, but now contains a searchable, selectable text layer. Test it by pressing Ctrl+F in your PDF viewer.

Frequently Asked Questions

How accurate is PDFusion's OCR?

PDFusion uses Tesseract OCR, one of the most accurate open-source OCR engines available. For clean, well-scanned documents in standard fonts, accuracy is typically 95–99%. Accuracy decreases with poor scan quality, unusual fonts, or handwritten content.

What languages does the OCR support?

PDFusion's OCR is optimised for English. Documents in other Latin-script languages (French, German, Spanish, Italian, Portuguese, etc.) will generally work well. Non-Latin scripts (Arabic, Chinese, Japanese, Korean, etc.) are not currently supported.

Does OCR change how the document looks?

No. The standard OCR output preserves the original scan appearance exactly. The text layer is added invisibly beneath the scan image — the document still looks like your original scanned pages, but the text is now searchable and copyable.

Can I run OCR on a PDF that already has some searchable text?

Yes, but it's usually unnecessary for pages that are already digitally created. If a PDF has a mix of scanned pages and digitally created pages, OCR is applied to all pages. On already-digital pages, OCR may produce a duplicate text layer, which is harmless but redundant.

Will OCR work on handwritten notes?

OCR accuracy on handwriting varies significantly. Neat, printed block-capital handwriting is often recognised well. Cursive handwriting is much harder to recognise reliably. For critical content in handwritten documents, always verify the OCR output carefully.

Make your scanned PDF searchable

Free OCR — no sign-up, no file storage.

Run OCR →

See also: PDF to Word · Compress PDF · All PDF Guides

Save result to cloud Save your processed file directly to Drive or Dropbox