A PDF is a container, not a promise of editable text
PDF is designed to preserve the appearance of a page. One file may contain letters, fonts and text positions; another may simply hold a photograph of a page. When you view them side by side, both look like documents, but only one necessarily has text a computer can read directly.
What is a selectable-text PDF?
A PDF exported from a word processor, presentation application or many digital forms usually stores actual characters. You can often highlight a word with your cursor, search for a phrase and paste the text into another application. Extract Text Now reads this text data directly, which is usually quick and more accurate than recognition from images.
Even these PDFs can be awkward. A page may have multiple columns, text inside shapes, headers that are stored after the body or fonts with unusual character mappings. If a copied result looks strange, the text storage order might be the cause.
What is a scanned PDF?
A scanned PDF commonly contains an image created by a document scanner or phone camera. The text is visible to a person, but a normal extraction program sees pixels rather than letters. You might be able to search the filename but not the printed words inside the page.
How optical character recognition works
OCR software examines an image, identifies patterns that resemble lines, letters and spaces, and guesses the words they form. Extract Text Now offers optional browser-based OCR using Tesseract for English text on pages without a meaningful selectable text layer. Because it must examine the picture, OCR takes more computation than direct extraction.
Five ways to improve OCR results
- Use a clear scan. Low resolution makes letters blur together. If you can rescan, choose a readable document setting.
- Keep pages straight. Rotated or curved pages make it harder to identify lines correctly.
- Improve contrast. Dark text on a light, clean background is easier to recognise.
- Crop unrelated background. Desk edges, shadows and other objects may confuse recognition.
- Proofread important details. OCR cannot guarantee exact names, quantities, dates or punctuation.
Which approach should you choose?
Start with direct extraction whenever you can select words in the original. Enable OCR when the page is image-only or extraction produces no meaningful text. For a document that mixes normal PDF pages and scans, Extract Text Now can attempt direct extraction first and use OCR only on the nearly empty pages.
Try a small page range to compare results before processing the entire document, and consult our formatting troubleshooting guide when the reading order is unexpected.