How to make a scanned PDF searchable with OCR

You open a scanned document, press Ctrl+F, type a word you can plainly see on the page — and get nothing. No results. The word is right there.

This is the single most confusing thing about scanned PDFs, and the explanation is simpler than it seems: to your computer, that page contains no words at all.

A scan is a photograph of paper

When a scanner or a phone camera captures a page, it records a grid of coloured dots. That grid, to a human eye, resolves into letters and sentences. To software, it is an image — the same kind of object as a photograph of a beach. There is no text in the file, so there is nothing for a search to find.

This is why several things fail in ways that seem inexplicable:

• Ctrl+F finds nothing, even for words clearly visible on the page.
• You cannot select or copy a paragraph with your mouse.
• Converting the file to Word produces an empty document.
• A screen reader has nothing to read aloud to a blind user.
• Extracting the text returns nothing at all.

What OCR does

Optical character recognition analyses the image, identifies the shapes that are characters, works out which characters they are, and writes them into the file as a real text layer positioned invisibly behind the picture.

The visual result is unchanged. Your page still looks exactly like the scan it always was. But now there is text underneath it, aligned to the words in the image, and every piece of software that previously found nothing suddenly works.

OCR does not change how your document looks. It changes what your computer can do with it.

Why this matters more than it sounds

The clearest payoff is in archives. A folder of two hundred scanned invoices is nearly useless — finding the one from a particular supplier means opening files one at a time. Run OCR across them and the whole archive becomes searchable in seconds.

Legal teams run OCR over discovery bundles so a name can be traced across thousands of pages. Researchers do it to digitised journal articles. Accountants do it to receipts. And there is an accessibility dimension that is easy to overlook: a screen reader cannot read an image of text, so an un-OCR'd scan is completely inaccessible to blind users. In many jurisdictions, publishing documents in that state is a compliance problem.

How to run OCR on a PDF

1. Upload your scanned PDF.
2. Wait while each page is analysed and its characters are recognised — this is genuinely slower than other PDF operations.
3. Download the new file, which looks identical to the original.
4. Open it and press Ctrl+F, then search for a word you can see on the page.
5. If it is found and highlighted, the text layer is working.

That last step is worth doing every time. It takes five seconds and confirms the recognition actually succeeded, rather than assuming it did and discovering otherwise when you need the document.

Getting accurate results

OCR accuracy depends almost entirely on the quality of the scan you feed it. On a clean 300 DPI scan of ordinary printed text, accuracy is typically well above 98%. On a faint fifth-generation photocopy, it can be close to useless.

• Scan at 300 DPI. This is the sweet spot. Below 200 DPI, characters blur into one another and error rates climb sharply. Above 400 DPI you gain very little and the files get large.
• Straighten the page first. Recognition assumes roughly horizontal lines of text, so a page scanned at an angle produces noticeably worse results. Fix rotation before running OCR, not after.
• Prefer the original over a photocopy. Each generation of copying degrades the characters, and OCR is far more sensitive to that degradation than your eye.
• Increase contrast on murky scans. Grey text on a grey background is hard for recognition; pushing it towards black on white helps considerably.

One firm limitation: handwriting will not work. OCR recognises printed type. Handwritten notes, signatures and annotations produce nonsense rather than partial results, and no amount of scan quality changes that.

The order to do things in

If your goal is an editable Word document from a scan, the sequence matters. Converting a raw scan directly to Word produces an empty file, because there is no text to convert. Run OCR first, then convert the searchable output. This catches people out constantly and the fix is simply doing the two steps in the right order.

Similarly, if you want to compare two scanned documents, or extract their tables into a spreadsheet, OCR both files first. Every text-based tool needs text to work with.

Why OCR takes longer than other tools

Most PDF operations rearrange existing data. Merging copies pages; rotating changes a flag; splitting rewrites an index. OCR has to render every page, analyse its pixels, and match shapes against character models. That is genuinely heavy computation.

A long document taking a minute or more is normal rather than a sign something has stalled. It is worth the wait for what you get back: a document your computer can finally read.