How to make a scanned PDF searchable

You press Ctrl+F in a scanned PDF and get nothing, even though the word is right there on the page. That's because the page is a picture of text, not text. Here's how to fix it properly, and how to check the result before you rely on it.

Why search finds nothing

When a scanner or a phone camera turns a page into a PDF, it saves a photo. The PDF is just a wrapper around that photo. To you it looks like text. To the computer it's a grid of coloured dots, and there's nothing in it to search, select or copy.

There's a quick way to tell. Try to select a single line with your mouse or finger. If nothing highlights, or the whole page gets selected as one block, you're looking at an image.

What OCR actually does

OCR stands for optical character recognition. The software looks at the shapes on the page, works out which letters they are, and writes the words into an invisible layer placed exactly behind the image. The page still looks the same, but now there's real text underneath it.

That one change unlocks a lot: search works, you can select and copy a sentence, highlights snap to words, and screen readers can read the page aloud. Keep one thing in mind, though. OCR is an educated guess. A good scan gets a very good guess. A poor one gets errors you won't see until you search for a word and it isn't there.

Get a clean scan first

The quality of the result depends far more on the scan than on the software. Before you run OCR:

• Keep the page flat and straight. Curved lines near a book's spine and tilted pages cause most recognition errors.
• Use even light. The shadow of your phone or hand across the page is enough to lose a few lines.
• Make it sharp. Hold still and tap to focus. Blurry letters become wrong letters.
• Use enough resolution. Around 300 dpi is the usual recommendation for ordinary text. Much lower and small print starts to break up.
• Crop the edges. Dark borders, fingers and the edge of the next page confuse the layout detection.

What OCR struggles with

• Handwriting. Printed text works well. Handwritten notes usually don't.
• Columns and tables. The words get recognized, but the reading order can jump between columns or across table cells.
• Tiny print and faint photocopies. Footnotes and copies of copies lose detail.
• Marks over text. Stamps, underlining and highlighter streaks can break a word in two.
• Formulas. Maths and chemistry notation rarely comes out right.

Also make sure the language setting matches the document. OCR uses its knowledge of the language to decide between similar-looking letters, and accented characters suffer most when it's set wrong.

Check the result before you trust it

A file that "has OCR" can still be full of holes. Two minutes of checking saves you from trusting a search that misses half the matches:

1. Search for three or four words you can clearly see, on different pages.
2. Copy a full paragraph into a plain text editor and read it next to the original.
3. Look out for the classic swaps: "rn" read as "m", "l" as "1", "O" as "0", missing accents.

If a key term is badly recognized, search for a shorter piece of it. "lastic" finds both "elasticity" and a damaged "e1asticity".

A worked example

Jonas photographs a 30-page library chapter with his phone, page by page, and turns the photos into one PDF. When he searches for "elasticity", nothing comes up.

He runs OCR. Now the search finds the word on most pages, but not on two pages he shot under a desk lamp, where his phone threw a shadow across the text. He retakes those two pages in daylight and runs OCR again. They're fine.

One page still misses a match. A highlighter streak from a previous reader runs through the line, and copying it shows "e1asticity". He can't fix the paper, so he searches for "lastic" from then on. The whole chapter is now searchable, and he knows exactly where its weak spot is.

Where your file goes matters

Many free OCR websites work by uploading your file to their server and sending back the result. For a public article, that's fine. For a contract, a payslip, a medical letter or a scan of your passport, it's worth knowing where the file ends up before you send it.

That's why ClearRead's PDF tools, OCR included, run on your own device. The file is never uploaded, there's no daily limit, and the page looks exactly as it did, with the text layer added underneath.

Try it on one file

Pick the scanned PDF you search through most often. Check whether it already has text, run OCR if it doesn't, and test it with three words. From then on, finding something in it takes a second instead of ten minutes of scrolling.