Is your PDF scanned or real text? How to tell, and why it matters

We build browser-based file tools and test every one of them against the forms and specs people actually have to satisfy.

The PDF to Word converter produced an empty document. The search box finds nothing although the word is clearly on the page. Copy and paste gives you nothing at all. The file looks completely normal.

Almost certainly, your PDF is a picture of a page rather than a page. This single distinction explains most of the moments when a PDF tool appears to be broken.

Two files that look identical

A digital PDF is created by software — Word, a browser, an invoicing system. It contains text as text: character codes, fonts, positions. The computer knows that a particular shape on page three is the letter g.

A scanned PDF is created by a scanner or a camera. It contains one photograph per page wrapped in a PDF envelope. There is no text anywhere in the file — only an arrangement of dark and light pixels that your brain reads as writing. The computer has no idea what is on it.

On screen they can be impossible to tell apart, which is why the failure is so confusing. The difference only appears the moment something tries to use the text.

How to tell, in about five seconds

Try to select a word. Open the PDF in any viewer and drag across a line of text.

• Individual words highlight, with the selection following the line → digital text.
• You get a rectangle over part of the page, or nothing at all → scanned image.

Then try searching. Ctrl+F for a word you can see in front of you. No results means no text layer.

There is a third case worth knowing about: a scan that has already been through OCR. It is still a picture, but with an invisible text layer laid over it. Selection works, searching works, and copying gives you text with occasional odd characters — rn read as m, 0 and O swapped. That tells you the text is a machine’s best guess rather than the original.

The scan checker answers this directly for a whole document, page by page, which is quicker than selecting your way through forty pages — and it tells you which pages have text, which matters for files that mix both.

Why files end up mixed

A single PDF often contains both kinds of page, and this catches people out. A contract generated digitally, printed for signature, signed, scanned and appended produces a file where pages 1–4 are real text and page 5 is a photograph. Conversion then half-works, which reads as a bug.

What each kind can actually do

The middle rows are the useful ones. Page operations work on anything, because they move pages around without looking inside them. Everything involving words needs a text layer.

Turning a scan into something usable

OCR — optical character recognition — looks at the picture and works out what the letters are, then writes a text layer into the file. The page still looks exactly the same; it just becomes searchable, selectable and convertible.

OCR a PDF does this in your browser, which is unusual: OCR is normally a server job, and a scanned document is often precisely the thing you would rather not upload — a contract, a medical letter, an ID. The recognition engine runs on your device instead.

What decides whether OCR works well:

• Resolution. 300 DPI is the usual target. Below 200 accuracy falls off sharply.
• Straightness. A page photographed at an angle recognises badly. Scan flat.
• Contrast. Crisp black on white beats a grey photocopy of a photocopy.
• Typeface. Printed text is reliable; handwriting is effectively not readable by ordinary OCR.
• Language. Our OCR is English only, and accented or non-Latin scripts need a model trained for them.

Expect very good rather than perfect. Always read the output before relying on it — a digit misread in an invoice total is worse than no text at all.

Choosing the right route

• Scanned, and you need the words → OCR, then convert.
• Digital, and you want an editable document → PDF to Word directly.
• Digital, and you only want the text → extract text, which is faster and cleaner than a Word round-trip.
• Scanned, and you only need to rearrange pages → no OCR needed at all.
• Scanned and enormous → compression will help far more here than on a text PDF.

The short version

Try to select a word. If you cannot, the file is a picture and no amount of converting will find text that is not there — run OCR first. Page operations work on anything; anything involving words needs a text layer. And if a tool seems broken on one file and fine on another, this is almost always the difference.