Edit a scanned PDF: change the words on a scan, then make it searchable

Why nothing selects on a scan

A scanner or a phone camera makes a picture of the paper. The PDF holds that picture and nothing else: no letters, no words, no lines. Click on a word and the whole page highlights, because the whole page is one image. Search finds nothing, and copy gives you nothing.

You can check this in a few seconds. Drop the file on Inspect: our test scan of a lease shows “Pages with text: 0 of 1 checked”, “Scanned pages (image, no text): 1” and “Fonts: none (no text)”.

OCR (optical character recognition) reads the picture and writes the words it finds into an invisible layer on top of it, each word where the scan shows it. That makes the page searchable and copyable. It does not turn the picture into editable text, and that is the part most guides skip.

What happens when you try to edit OCR text

We ran our lease scan through OCR PDF and opened the searchable copy in the editor. Here is what we found, so you do not have to find it the slow way:

The last row is the method. It is the honest version of white-out, and on a scan it is the right tool: there is no text in the picture to edit, so covering the ink and writing new text is all any editor can do without redrawing the scan.

Step 1: cover the old word and type the new one

Start from the scan as it came, before any OCR. If you only have a searchable copy, see the note at the end of this section.

1. Open the scan in the editor above, or at /edit. Zoom in on the word to change.
2. Press S for the Shape tool and pick the rectangle. Set the line colour and the fill to white.
3. Drag a rectangle over the old word. Select it and nudge with the arrow keys (1 point a press, 10 with Shift) until no ink shows at its edges.
4. Press T for the Text tool. Choose Sans or Serif to match the scan, set the size, and pick a dark grey rather than pure black if the scan’s ink is soft. On our lease, 9 pt Sans matched the printed figures.
5. Click on the white and type the new word. Press Enter to place it, then drag it until it sits on the same line as its neighbours.
6. Save with Ctrl+S. The file downloads with “-edited” added to its name.

New text is set in Helvetica, Times or Courier, the fonts every PDF viewer has. A scan has no font to match, so aim for “the same size and weight”, not “identical”. On a printed form or letter that is close enough to pass; on a page of handwriting it never will be, and nobody should pretend otherwise.

If your file already went through OCR, its invisible layer still holds the old word under your white box, and search and copy will find it. Running OCR again on top would read every word a second time. Where you can, go back to the original scan; where you cannot, accept that the old word stays findable, or use Redact for it (next sections).

Step 2: run OCR on the edited file

1. Drop the edited file on OCR PDF above, or open /ocr-pdf.
2. Pick the Language the document is written in. A second language under And helps a page that mixes 2, and is slower.
3. Set Pages with text to read again too. Your new word counts as text, so by default OCR skips the page and says: “every page already has text, so nothing was added”.
4. Run it and read the report: how many words were made searchable, the average confidence, and any words it was unsure of.
5. Download the copy with “-searchable” in its name.

On our edited lease, OCR read 27 words at 95 % average confidence in about 1 second. Search in the result finds “1300” and no longer finds “1250”. It finds “1300” twice, once as the text you typed and once in the recognised layer, which is harmless.

OCR runs on your device too: the engine and the language data download once, the first time you use the tool, and then the page works without a connection. Nothing is uploaded.

What the result leaves in the file

Be clear about what you made:

• The page shows the new word. It prints and displays the same in every viewer.
• The old word is still in the scan picture, under a white rectangle. The rectangle is a separate object: anyone who opens the file in an editor can move it and read what is underneath.
• Search and copy find the new word, if you ran OCR last as above.

That is fine for a typo in your own letter or a corrected date on a form you are about to send. It is not fine for a number that must not be readable at all, like an account number or a salary. For those, use Redact PDF: press R, drag a box over the word, and save. Redaction paints over the scan’s pixels under the box and takes out any recognised text there, so nothing is left to uncover. The box is black, not white, which is the point: a reader can see that something was removed.

The white-out guide compares covering, editing and redacting in more detail, measured on our invoice test file.

When to stop and retype instead

Covering and retyping works for a word, a number or a date. It stops being worth it when:

• The change is a sentence or more. Matching the scan’s spacing line after line takes longer than retyping the page. Run OCR, then PDF to Word to get the text into a document you can edit freely; the layout comes out approximate, and the tool says so before you start.
• The page is handwritten. Your typed word will look typed. Our OCR measurements on handwriting-style fonts ran from 72 % to 94 % confidence, with some digits misread, so the searchable layer needs checking too; the handwriting guide has the numbers.
• The scan is skewed, blurred or tiny. OCR misreads, and a straight text box on a tilted line looks wrong. Rescanning at 300 dpi, flat and straight, saves more time than fixing a bad scan.
• The document is official. A certificate, a signed contract or a bank statement should be corrected and reissued by whoever issued it, not patched.

For everything about OCR itself (languages, accuracy, phone scanning habits), see the scanned PDF guide. For editing PDFs that have real text, start with the editing guide.

How other tools handle it

Adobe’s help page “Edit scanned PDFs” (checked on 11 October 2026) says Acrobat runs OCR in the background when you open a scan for editing, turns it into editable text and images, and tries to match the look of the original fonts. That is a different approach from ours: it redraws the text it recognised rather than covering the ink, which reads more smoothly when it works and depends on the OCR being right when it does not. getPDF does not redraw scans; it keeps the picture, lets you cover and retype, and makes the result searchable. If you need many paragraphs changed on a scan, a tool that rebuilds the page, or retyping, is the better choice.