← All posts

Why is my PDF text not selectable?

You can see the words. You drag the cursor across them and nothing highlights — or something highlights, and what lands in the paste is empty, or is þÿßæ. Either way the document looks like text and behaves like a photograph.

This has at least four different causes. They are not variations of the same problem, and they need completely different fixes, so the first thing worth doing is finding out which one you have. It takes about thirty seconds and it stops you paying for a converter that was never going to help.

Every page that answers this question tells you to run OCR. OCR is the right answer to one of the four.

Why a PDF can show text it will not let you select

A PDF page is a list of drawing instructions, not a document in the sense a word processor means. The instructions say things like "set font F1 at 11 points, move to this coordinate, draw the glyphs with these codes". Selecting text means the reader has to run that backwards: find the glyphs, work out which characters they stand for, and guess which ones belong to the same line.

There are four places that can break, and each one breaks it differently.

There are no text instructions at all. A scan, a photograph or a "print to image" export has one instruction on the page: draw this picture. The words are pixels inside it. There is nothing behind them to select, and no amount of clicking will produce any, because the information does not exist in the file.

The text is there and the reader is refusing. PDFs carry permission flags — may print, may copy, may edit. A file can be encrypted with an owner password and an empty user password, which means it opens with no prompt at all, looks entirely normal, and quietly disables the copy command. Nothing is wrong with the file. A compliant reader is doing what it was asked.

The glyphs have no character codes. A font maps character codes to glyph shapes. Going the other way — from the glyph the page drew back to the letter it represents — needs a ToUnicode table, and that table is optional. Subset a font on export, or use a badly written generator, and you get a document that draws perfectly and has no idea what any of it says. That is the gibberish case: the reader hands you the raw codes, and they paste as accented rubbish or as boxes.

The words are not on the page. Form field answers and sticky notes are separate objects floating above the page, drawn by the reader on top of it. The text-selection tool works on page content, so it sails straight past them. This is why a completed form can look full of typed answers and extract as an empty document.

Working out which one you have

  1. Try to select a single word. If nothing highlights anywhere on the page, you are in cause 1 or cause 4. If a blue selection appears, the text is real and you are in cause 2 or 3.
  2. If nothing highlighted, click directly on a filled-in answer in a box or a line. If the cursor lands in it and you can retype it, those are form fields — cause 4. If the whole page acts as one object, it is a picture — cause 1.
  3. If it highlighted but copy did nothing, look at the reader's Edit menu. A greyed-out Copy, or a padlock in the title bar, is cause 2.
  4. If it copied but pasted as nonsense, that is cause 3, and it is a font problem rather than a file problem.
  5. Confirm it against the file itself. Our PDF inspector reads page count, page sizes, encryption and whether there is a real text layer, entirely in your browser — the file is never uploaded. For a long document, OCR PDF goes further and tells you page by page which pages carry real text, which are pictures, and which are simply blank.

That last one matters more than it sounds. "My PDF is not searchable" is usually true of three pages out of two hundred — the scanned appendix somebody stapled onto a normal report — and knowing which three is the difference between a five-minute job and re-scanning a folder.

Cause 1: it is a picture

The only fix is optical character recognition: software looks at the pixels, decides what letters they are, and writes an invisible text layer underneath the image. After that, selection and search work, because now there is something to select.

Run recognition only over the pages that need it. Re-recognising a page that already has good text lays a second, guessed copy of the words underneath the real one, and every search then matches twice. If your next step is a Word document, the order of operations matters and it is not the obvious one.

Cause 2: the copy command is blocked

The file opens without asking for anything, so people assume it is unprotected. It is not — it is encrypted with an empty user password, and the copy flag is off.

Two routes. Locally, qpdf --decrypt in.pdf out.pdf strips the security dictionary and never sends the document anywhere, which is what you want for anything confidential. Or use our Unlock PDF tool and leave the password box empty: a restrictions-only file needs no password, because the document was never closed to you in the first place — only its flags were set. If it does ask for a password you do not have, nothing here or anywhere else gets you in; that is what encryption means.

It is worth knowing what those flags actually are if you ever set them yourself, because password protection is two different things in one dialog. The encryption is real. The permissions are a request that Acrobat, Preview and the browser viewers honour and other software need not — they stop the copy command, not a screenshot, and not an extractor that does not ask.

Cause 3: it copies as gibberish

The page draws correctly and pastes as nonsense, so the glyphs are there and the character map is missing or wrong. Nothing can reconstruct meaning that was never written into the file — but before you conclude that, check whether the problem is your reader rather than the document. Put the file through PDF to text: if the extracted text is clean, your reader's copy is at fault and you have your text. If it comes out as the same rubbish, the map really is broken.

From there:

Cause 4: the words are in form fields

Flattening presses those floating objects down into the page. Flatten PDF draws each field's appearance into the page's own content and deletes the field, so the answers become ordinary page text — selectable, searchable, and included by any extractor. It is also the standard step before a completed form is filed, since nobody can re-type a figure afterwards.

Keep the original. There is no unflatten, the form can never be filled in again, and if what you actually wanted was the answers without the fields that is the same operation described from the other end.

The accessibility angle

A screen reader may happily read aloud a document you cannot select a word of, which sounds like a contradiction and is not. Assistive technology has an explicit exception in the permissions model: reading aloud stays allowed whatever the copy flag says. And a tagged PDF carries a structure tree describing headings, lists and reading order — a screen reader can walk that even where the visible selection behaves oddly.

The reverse is more common and worse. A scan is silent to a screen reader, invisible to every search index, and a straight failure of any accessibility check. If the document has to be accessible, cause 1 is not an inconvenience — it is the document not working.

What none of this will do

Text you cannot select is sometimes text that was never captured. A photograph of a whiteboard, a fax that went through twice, a 200 dpi scan of a carbon copy — recognition will produce something from all of those, and what it produces may be worse than nothing, because it looks like text and is wrong. Check the first page of any recognised output before you trust the other four hundred.

And nothing here puts information into a file that never had any. Our OCR detects and does not recognise, the editor cannot type into a picture, and PDF to text stops and says so rather than handing you an empty file from a scan. Knowing which of the four causes you have is worth more than any single tool, which is why it is the first thing on this page.

Tools this post uses: