How to prepare a scanned PDF for more accurate OCR

Understand what OCR can and cannot recover

Optical character recognition estimates characters from pixels. It can make a clean scan searchable and reusable, but it cannot restore detail that the scanner never captured. Small text, compression blocks, shadows and handwriting all reduce certainty.

A text-based PDF usually gives a better result than a scan. Before using OCR, try selecting a sentence in your PDF viewer. If the words can be selected in the correct order, use the original file rather than printing and rescanning it.

Capture a clean page

For office documents, a scan around 300 dpi is usually a useful balance between clarity and file size. Photographing pages can work, but a flat scanner or a document-scanning mode is more reliable because it controls perspective and lighting.

Keep the entire page inside the frame. Cropped dates, account columns and footnotes can change the meaning of extracted content. Remove fingers, desk edges and strong shadows before saving the PDF.

Review checklist

• Use approximately 300 dpi
• Keep all four page edges visible
• Avoid glare, shadows and perspective distortion

Fix rotation, contrast and page order

Rotate every page upright before conversion. Mixed orientation forces the OCR engine to detect direction repeatedly and can produce inconsistent reading order. Very pale scans benefit from improved contrast, while aggressive black-and-white filtering can erase decimal points and thin table lines.

Put pages in their final order and remove accidental duplicates. For double-sided scans, check that blank backs have not interrupted the document or pushed a heading away from the content it describes.

Review the risky parts first

Numbers deserve more attention than ordinary prose. Check account numbers, dates, monetary values, percentages and identifiers character by character. Common OCR confusions include zero and the letter O, one and lowercase l, or punctuation disappearing from decimal values.

Tables and multi-column pages also need a visual comparison. OCR may recognise every word but return them in the wrong reading order. A searchable PDF can preserve the original appearance, while an editable Word conversion has to reconstruct that structure.

Review checklist

• Check numbers and decimal separators
• Compare table columns
• Confirm headers and footers were not inserted into body text

Protect sensitive documents

Remove pages that are not needed before uploading, especially identity documents or unrelated account records. Use a trusted connection and review the service's retention terms. The File Workshop keeps uploaded documents separate from advertising data and publishes its retention controls in the Privacy and Security pages.