How to make a scanned PDF searchable (OCR)
Last updated: July 7, 2026
You press Ctrl+F, type the invoice number you know is in there somewhere, and get nothing. The document is right in front of you, the number is plainly visible on page 3, but as far as your computer is concerned the PDF contains no text at all. This is the everyday frustration of scanned PDFs: they look like documents, but they behave like photographs.
The fix is OCR, and these days you don't need Acrobat, a scanner suite or a subscription to run it. You can do it in your browser in a couple of minutes, and, if you use the right tool, without your document ever leaving your machine. Here's how, plus what OCR can and cannot realistically do for you.
Why scanned PDFs are dead documents
A PDF made from a Word file or a web page carries real text inside it: every letter is stored as a character, which is why you can search it, select it and copy it. A PDF made by a scanner or a phone camera carries no characters at all. Each page is one big picture, a photo of text rather than text itself. Search finds nothing, copy-paste grabs nothing, screen readers read nothing, and file indexers like Windows Search or Spotlight skip straight past it.
OCR, short for optical character recognition, is the technology that bridges the gap. An OCR engine examines the image, recognises the shapes of letters and words, and produces actual machine-readable text with the position of every word on the page. Modern engines are remarkably good on clean printed pages: the open-source Tesseract engine, originally developed at HP and later maintained by Google, sits behind many desktop OCR applications and routinely reads a decent office scan with very few errors.
Make a scanned PDF searchable, step by step
The robinpdf OCR tool runs a WebAssembly build of that same Tesseract engine directly inside your browser. The process takes a few clicks:
- Open the OCR PDF page.
- Click the box and pick your scanned PDF, or drag the file in.
- Choose the language of the document: English, Spanish, or both if the document mixes them. Picking the right language matters, because the engine uses it to resolve ambiguous letter shapes.
- Press Run OCR and let it work through the pages. Recognition happens on your own processor, so a long document takes longer on an older laptop, exactly as a desktop OCR app would.
- Download the result. It looks pixel-for-pixel identical to your original scan, but Ctrl+F now works and you can select and copy the text.
One detail worth pausing on: nothing is uploaded. The traffic actually flows the other way round. The OCR engine itself downloads to your browser the first time you use the tool, then reads your document locally. For a scanned contract, a medical report or a box of old bank statements, that is a meaningful difference from services that ask you to hand the file over first. If you want the longer discussion of what happens to files on conventional upload-based sites, we've written it up in Are online PDF tools safe?
How the invisible text layer works
A common worry is that OCR will mangle the look of the document, and with some tools that reconstruct the page from recognised text, it genuinely can. This tool takes the safer route used by professional archiving software: it keeps the original scanned image exactly as it is and places the recognised text underneath it, invisibly, with each word positioned where it appears in the image.
The result is often called a "sandwich PDF". What you see is still the original scan, stamps, signatures, coffee stains and all. What the computer sees is a full layer of text lined up with the picture. When you search, the match is found in the hidden layer and highlighted on the image above it. When you select and copy, you are copying from that layer too. Nothing about the visible page changes, which also means OCR never degrades the scan quality: the image is passed through untouched.
This is also why a freshly OCR'd file can be slightly larger than the original: it now holds the image plus the text. If size matters, run the result through Compress PDF afterwards; scans usually shrink a lot.
Where searchable scans really pay off
- Receipts and invoices. Scan them as they arrive, OCR them, and at tax time you can search a whole folder for a supplier name or an amount instead of leafing through pages.
- Old family documents. Typed letters, certificates and records become searchable and quotable, which makes genealogy work dramatically faster.
- Research papers. Older journal articles often circulate as raw scans. OCR lets you search them and copy citations instead of retyping quotes.
- Archived contracts. Finding one clause across a shelf of scanned agreements goes from an afternoon to a keystroke. If a contract contains details you need to share carefully, pair OCR with Redact PDF before passing it on.
Getting the best results from your scans
OCR accuracy depends far more on the input than on the engine. A few habits make a large difference:
- Aim for 300 DPI. It is the long-standing rule of thumb for OCR. 200 DPI often still works; below that, letters lose the detail the engine needs.
- Keep pages straight. A page scanned at an angle recognises noticeably worse. Most scanner software has a deskew option; use it.
- Light it evenly. If you photograph documents with a phone, avoid shadows across the page and shoot from directly above.
- Prefer black on white. Coloured paper, highlighter marks and faint toner all cost accuracy.
- Set the language correctly. Running an English model over Spanish text produces plausible-looking nonsense around accented characters.
- OCR one combined file. If your scan arrived as separate page files, merge them into one PDF first, then OCR the lot in a single pass.
Honest limits: what OCR won't do
It is worth being straight about the edges. OCR engines are built for printed text, so handwriting is largely out of reach: a handwritten letter or a filled-in form field will come out as gibberish or nothing, and no amount of rescanning fixes that. Low-quality scans, faxed documents, blurry phone photos and tiny print will produce errors, sometimes many. And while the words in a table are recognised, the table's structure is not: copy a table out of an OCR'd page and you get a stream of cell values, not rows and columns you can paste into a spreadsheet.
None of this makes OCR less useful for its main job. Even an imperfect text layer with a few misread characters still lets you find the page you need in seconds, which is usually the whole point.
Frequently asked questions
Will my PDF look different after OCR?
No. The visible pages are your original scan images, unchanged. The recognised text is added as an invisible layer beneath them.
Is my document uploaded anywhere?
No. Recognition runs inside your browser on your own device. The engine downloads to you; your file never leaves your machine.
Which languages are supported?
English, Spanish, or both together for documents that mix the two.
Can it read handwriting?
Not reliably. OCR engines are trained on printed type. Neat block capitals occasionally come through; cursive does not.
Why is OCR slow on my machine?
Because the work happens on your processor rather than a server. A modern laptop handles a typical document quickly; a long scan on an old machine takes a while. It only needs doing once per file.
The output file got bigger. Is that normal?
Yes, slightly: it now contains the images plus the text layer. Run it through Compress PDF if you need it smaller.