How to make a scanned PDF searchable

Why your scan isn't searchable

PDFs hold two very different kinds of “text”. A file made from Word contains real text — characters the computer knows are characters, which is why search, copy and screen readers work. A scanner produces something else entirely: a picture of the page. To a computer, that picture is millions of coloured dots; the fact that the dots form words is invisible to it. Search finds nothing because, as far as the file structure goes, there is nothing to find.

What OCR actually does

OCR — optical character recognition — is software that looks at the picture the way a person does: it finds line shapes, cuts them into character shapes, recognises each one, and assembles words. The clever part is what happens next. Good OCR does not replace your scan. It writes the recognised text into an invisible layer positioned exactly behind the printed words. The page still looks like the original scan, pixel for pixel — but now, when you search for a name, the search engine finds it in the hidden layer, and when you select and copy, you get real text. One file, both natures. This is what “searchable PDF” means.

What decides whether OCR works well

• Scan quality beats everything. 300 DPI, flat page, even lighting. Most “OCR is bad” complaints are actually “my source photo was crooked and shadowed” complaints. A straight, well-lit capture at decent resolution routinely reads at 98–99% accuracy for printed text.
• Language matters. OCR engines use language models to disambiguate similar shapes (l, 1, I; O, 0). Telling the tool the document's language, especially for non-Latin scripts — Bengali, Arabic, Chinese — changes results dramatically.
• Printed vs handwritten. Printed text is a solved problem. Handwriting is not: expect real errors on anything handwritten, and proofread accordingly.

The three OCR jobs, and which you want

1. Make the scan searchable — add the invisible layer, keep the file looking identical. Right choice for archiving: contracts, records, anything where the visual original matters.
2. Extract the text — pull the words out into plain text you can paste elsewhere. Right choice when you need a quote, an address, a table's contents — the scan itself was never the point. Our Get text tool does this, with OCR kicking in automatically when the page turns out to be a scan.
3. Reconstruct the document — turn the scan into editable Word. The hardest job of the three, covered in our conversion guide.

A privacy note that matters more with OCR than anywhere else

Think about what people OCR: identity documents, medical results, old legal papers. Cloud OCR services read the full text of the document on their servers — that is the deal, stated or not. For sensitive material, prefer OCR that runs on your own device. Our web app's OCR runs in your browser; the Android app reads text on the phone itself. The document's contents never cross the network just to become searchable.

Practical checklist

• Scan at 300 DPI, pages flat, lighting even — or use a scanner app with edge detection and de-skew.
• Set the correct language before running OCR.
• Archiving? Make it searchable (invisible layer). Quoting? Extract text. Editing? PDF→Word.
• Spot-check names and numbers — the words a human would never misread are exactly the ones OCR sometimes does.
• Sensitive document? Use on-device OCR, not a cloud upload.

Keep reading