Text-based vs scanned PDFs

Two files can look identical on screen and be completely different inside. Only one of them can have its text edited.

By EasyEditPDF · Updated Sep 7, 2026

Telling them apart in ten seconds

Open the file in any viewer and try to select a line of text with the mouse. If words highlight, there is a text layer. If the whole page highlights as one block, or nothing does, you are looking at a picture of a document.

What each one is

A text-based PDF stores instructions: draw this string, in this font, at this position. The characters are really in the file, which is why they can be searched, copied and — with the caveats in how editing works — changed.

A scanned PDF stores one image per page. There are no characters anywhere in it. What looks like a letter is an arrangement of pixels, and nothing in the file records which letter it was meant to be.

How this service decides

Each page is measured rather than guessed at. A page is treated as scanned when it carries almost no text of its own — under half a percent of the page area — or when more than half its glyphs have no declared character, and images cover at least half the page. A page with a substantial image but real text alongside it is treated as mixed, and its text remains editable.

Both halves of that test matter. A page that simply has very little text on it — a title page, a signature page — is not called scanned, because there is no image dominating it.

A scan someone has already put a text layer on

If a file has been through character recognition elsewhere, it carries a text layer sitting on top of the page image. This service reads that text layer like any other, so the file is not treated as scanned. What usually follows is that the text sits over a photograph, so there is no flat colour behind it — and blocks in that position are graded overlay only.

What happens here with a scan

You can upload it, view it and download it. You cannot edit its text, and this service will say so rather than letting you try. It does not perform character recognition, and there is no setting, plan or queue that turns that on.