Files are processed in your browser, not uploaded. Privacy →
Deckle
Open a PDF
Skip to guide

Extract text

Native PDF text vs OCR text: what is the difference?

By DeckleUpdated

Native PDF text is the exact characters the creating software wrote into the file, such as Word or a browser saving to PDF. OCR text is a guess: software looked at an image of the page, recognised the letter shapes and added an invisible text layer behind the picture. Both can be selected, searched and exported with PDF to text, but only native text is reliably correct. OCR text needs checking wherever names, numbers or dates matter.

PDF to text

Pages become plain text

Selected PDF pages

Order chosen in the review

UTF-8 text

Words without page formatting

Page boundary: U+000C or a blank line

Conceptual illustration · page separators retain boundaries, not columns, fonts or table geometry.

How the two kinds of text get into a PDF

Native text OCR text
Where it comes from The document's own characters, written at export Recognised from a scan or photo
Accuracy Exactly what the author typed Usually high on clean scans, with occasional misreads
What you see The text itself, drawn with fonts A picture of the page; the text sits invisibly behind it
Zoomed in Letters stay sharp Letters turn blocky or blurred
Selection Follows the letters closely Can be slightly offset from the visible words
Typical sources Word, Google Docs, browsers, invoicing and banking systems Scanners, phone scanning apps, OCR tools

A page with neither is image-only. It looks like text but holds no characters at all, so selecting, searching and extraction find nothing.

How to tell which one a page has

  1. Zoom in to 400% or more. Native text stays crisp; a scanned page shows pixels, speckles or a slightly grey background.
  2. Drag across a line. On an OCR page the highlight boxes may not line up with the words exactly.
  3. Search for a word you can see. If it is found, the page has text of some kind. If not, it is image-only.
  4. Extract the page with PDF to text and read the result. Odd substitutions such as c1ient for client point to OCR.
  5. Check the PDF's properties in a viewer. A producer such as a scanner or OCR application is a strong hint.

Example: a mortgage offer arrives as a 14-page PDF. Pages 1–12 have crisp text exported from the lender's system. Pages 13–14 are signed pages that were scanned and run through OCR. Extraction gives perfect text for the first twelve pages and a version of the last two with a few misread characters in the names and dates.

Common OCR mistakes to look for

OCR reads shapes, so it confuses characters that look alike, particularly on faint, small or skewed print.

  • 0 and O, 1, l and I, 5 and S
  • rn read as m, cl read as d
  • Commas and full stops swapped in numbers, such as 1.250,00 for 1,250.00
  • Accents dropped or added: e for é
  • Words split or joined where the scan had a fold, a stain or a stamp

When the text will be searched for an account number, a reference or a person's name, check those against the visible page.

What this means for text extraction

PDF to text reads whatever text layer the page has. On a native page it gives the author's exact text. On an OCR page it gives the recognised text, mistakes included; extraction copies them, it does not correct them. On an image-only page it gives nothing, and the result screen's Verified: line says there was no selectable text to compare.

So the question to ask before extracting is not just "is there text?" but "where did it come from?" For an image-only scan, the order is OCR first, then extraction. For a page with OCR text, extract it and proofread. For native text, extract and use. PDF to text vs OCR helps you choose.

Why you should not OCR a page that already has text

A page with native text already holds the author's exact characters. OCR can only produce a recognised copy of the same words, with the risk of misreads, so it adds nothing useful. If your OCR tool lets you choose pages, pick only the pages that are pictures.

Frequently asked questions

Is OCR text the same as real PDF text once it is added?

It behaves the same way: you can select, search and extract it. It is only as accurate as the recognition, so it can contain mistakes native text would not.

Can I see the OCR layer?

Not normally. It is invisible by design, so the page keeps its scanned appearance. Extracting the text with PDF to text is the easiest way to read it.

Does PDF to text run OCR?

No. It reads the text layer that is already there, native or OCR, and does not recognise text from images.

Can I remove a bad OCR layer?

Not with PDF to text, which only reads the layer. The cleanest fix is to run OCR again on the original scan, from before any OCR was applied, and proofread the new result.

Related tools

  • PDF to text to export a page's text layer and read it
  • Split PDF to separate scanned pages from native ones
  • PDF to PNG to save a sharp image of a page for checking

Sources