OCR a Scanned PDF: Make It Searchable and Copyable for Free
By Chinmoy Biswas · Updated , first published
You search a 60-page scanned contract for “termination” and get no results, even though you can see the word on page 12. The PDF is a stack of photographs of paper. Optical character recognition turns those pictures into text a computer can search, copy and convert, while leaving the pages looking exactly as they did. This guide explains what happens to the file, how to run OCR without uploading anything, and how to get results you can rely on.
What is a searchable PDF?
A searchable PDF is a scan with an invisible layer of recognised text placed exactly over the words in the image. You see and print the scan; you search, select and copy the hidden text.
OCR software looks at each page image, identifies the shapes of letters and words, and records what it read and where. The PDF format supports text drawn in an invisible rendering mode, which is how the layer stays hidden. That is why the page does not change appearance after OCR, and why a recognition error never shows on the page. It only shows when you search or paste.
To check whether a PDF needs OCR, try to select a sentence or press Ctrl+F (Cmd+F on a Mac) and search for a word you can see. If nothing highlights, or the whole page selects as one block, it is a pure scan.
How do you OCR a PDF for free without uploading it?
Use an OCR tool that runs in the browser. OCR PDF loads the recognition engine into the tab and reads the pages on your device, so the scanned document itself is never sent anywhere.
The engine is Tesseract, an open-source OCR project that started at Hewlett-Packard in the 1980s, was released as open source in 2005 and was then developed with Google's backing for over a decade. Since version 4 it uses a neural network that reads whole lines of text. Here it is compiled to WebAssembly, and the English model, about 3 MB, is downloaded from this site once.
- Open OCR PDF and choose the scanned file, up to 100 MB. Enter the password first if the file is protected.
- Leave the default to skip pages that already contain text, or tick Also OCR pages that already have text if their existing text is wrong.
- Press Make searchable. Each page is rendered at about 250 DPI and read in turn. Expect roughly 5 to 20 seconds per page on a laptop and longer on phones; a 20-page scan usually takes two to seven minutes.
- Download the searchable PDF and test it: search for a word you can see on a middle page.
How accurate is OCR, and how do you improve it?
On a clean, straight 300 DPI scan of printed text, modern OCR gets around 98 to 99% of characters right. Accuracy depends far more on the input than on the software, so the best fix is nearly always a better scan.
The Tesseract project's own guide to improving output quality says the same: it works best at 300 DPI or more, with straight lines of dark text on a clean background and no dark borders.
- Resolution: 300 DPI is ideal for normal print; below 150 DPI, small text becomes guesswork. Capital letters should be at least 20 to 30 pixels tall.
- Straight pages: skewed or curved pages (a book photographed open) produce broken lines. Rescan flat if possible.
- Contrast: dark text on a light, even background. Shadows across a page cause missing words.
- Borders: black scanner edges get read as characters. Trim them with Crop PDF before OCR.
- Language: the tool currently recognises English. Other Latin-script languages are partly recognised, but accents may come out wrong, and Chinese, Arabic and Cyrillic are not supported.
- Printed text only: handwriting, stylised fonts and text over photos are unreliable in any free OCR engine.
What 99% accuracy means in practice
A page of 400 words has roughly 2,000 characters. At 99% character accuracy that is about 20 wrong characters per page, often in the places that matter most: 0 and O, 1 and l, 5 and S, rn and m. Search still finds most words, but do not trust a pasted invoice total or account number without comparing it with the image.
OCR before converting to Word
PDF to Word reads the text layer of a PDF and has no OCR of its own. A scan has no text layer, so converting it directly produces an empty or image-only document. Run OCR first, then convert the searchable PDF. The same applies to PDF to Text.
Expect to proofread. The Word file is built from the recognised text, so every OCR mistake becomes a visible typo. Tables come through as lines of text, not as table cells. For a long scanned document that needs heavy editing, retyping the key sections from a clean OCR paste is sometimes faster than repairing a full conversion.
Built-in alternatives on each platform
Every major platform can now read text from images, but most of them only let you copy it. They do not save a text layer into the PDF.
iPhone and Mac
Live Text works on iPhone XS and later with iOS 15 or later, and on Macs running macOS Monterey or later. Open a scan in Photos, Files or Preview and select text straight from the image. It is excellent for grabbing a paragraph, and it handles several languages. It does not change the file, so the PDF still cannot be searched in other apps or by Spotlight.
Android and Google Drive
Google Lens copies text from any image on screen. For whole documents, upload the PDF to Drive, right-click it and choose Open with > Google Docs. Google's help page lists the limits: files of 2 MB or less, text at least 10 pixels high, and the page the right way up. Images are often dropped, and the document is stored in your Google account.
Windows
Windows 11's Snipping Tool has a Text actions button that recognises text in a screenshot and copies it. It works a screen at a time, so it suits a paragraph, not a 40-page file. Windows has no built-in way to write a text layer into a PDF.
| Method | Saves a searchable PDF | File leaves the device |
|---|---|---|
| OCR PDF in the browser | Yes | No |
| Live Text on iPhone and Mac | No, copy only | No |
| Google Drive: Open with Google Docs | No, makes a Google Doc | Yes, stored in Google Drive |
| Google Lens on Android | No, copy only | Image is sent to Google |
| Snipping Tool Text actions on Windows 11 | No, copy only | No |
Check what OCR recognised
To see the recognised text, open the searchable PDF, select all the text on a page (Cmd+A or Ctrl+A in most viewers), copy it and paste it into a notes app. Garbled lines show where recognition struggled, typically in tables, stamps and small footers. Those areas will not be found by search either, so do not rely on search alone to find every mention of a name or amount in a poor scan.
Troubleshooting
Pages were skipped
By default, pages that already contain text are left alone. Some scanners add a tiny text stamp, such as a date or a Scanned by line, which makes a page count as having text. Tick Also OCR pages that already have text and run the file again.
The selection does not line up with the words
On skewed pages the invisible text follows straight lines while the printed lines slope, so highlights drift towards the end of each line. Search still works. If alignment matters, rescan the page straight, or rotate sideways pages first with Rotate PDF, because text lying on its side is not read at all.
OCR is very slow or the tab reloads on a phone
Recognition is heavy work, and phones limit the memory a browser tab can use. Keep the tab in the foreground, close other tabs, and process long documents in parts of 20 to 30 pages with Split PDF. A laptop is several times faster.
Good uses for searchable PDFs
- Archives of receipts, statements and letters that you can find later by searching your computer. Spotlight and Windows Search both index the text inside PDFs.
- Contracts and leases where you need to jump to specific clauses.
- Accessibility: screen readers can read the text of a searchable PDF aloud; they get nothing from a pure scan. The output is not a fully tagged accessible PDF, but it is a large step up.
- Quoting: copy a passage instead of retyping it.
Size and privacy notes
The invisible text layer adds very little size, typically a few kilobytes per page, and the original page images are kept untouched. If the scan is large and has to meet an upload limit, run OCR first and then Compress PDF at Recommended, which keeps the text layer. Extreme would turn the pages back into plain images and discard it.
Because recognition happens on your device, OCR PDF is suitable for medical records, legal documents and anything else you would hesitate to upload to a cloud OCR service.
FAQ
Questions people also ask
Can I OCR a PDF for free?
Yes. OCR PDF is free with no sign-up and adds a searchable text layer to scanned pages in the browser. Google Drive can also extract text for free by opening the PDF as a Google Doc, but it does not return a searchable PDF.
How do I know if my PDF has already been through OCR?
Try to select a word. If single words highlight and Ctrl+F or Cmd+F finds a term you can see on the page, a text layer exists. If nothing highlights, or the whole page selects as one picture, the PDF is a plain scan.
Does OCR change how my PDF looks?
No. The scanned page images stay exactly as they were, and the recognised text is added on an invisible layer above them. Printing gives the same result as before. File size grows by only a few kilobytes per page.
Can OCR read handwriting in a scanned PDF?
Not reliably. Tesseract and other free engines are trained on printed type, so handwriting, signatures and filled-in form answers mostly come out as noise. Neat block capitals sometimes work. For handwritten notes, plan to retype the parts you need.
Free tool
Try OCR PDF
Recognize the text in scanned pages and add an invisible text layer you can search, select and copy, while the pages look exactly as before. Runs in your browser, so the file never leaves your device.
Open OCR PDF