How to Make a Scanned PDF Searchable with OCR, Step by Step

What “searchable” actually means

A searchable PDF from a scan has two layers on each page. The visible layer is an image of the paper. On top of it sits invisible text, placed word by word at the positions where the OCR engine found each word. When you search, select or copy, your PDF reader works with the invisible layer, so the highlight lands on the right spot in the picture.

That design has a consequence people miss. The text layer is only as good as the recognition. If “morning” is read as “moming”, a search for morning skips that page, even though the page image clearly says morning. That is why the scan quality numbers below matter more than which button you press.

Before you start: check what you have

• Try selecting a word in your PDF reader. If you can highlight individual words, the file already has text and OCR won't help. For fixing that text, use Edit PDF instead.
• Check the size. The OCR tool accepts PDFs up to 20 MB. A thick scanned file can be split into parts first and merged again afterwards.
• If the file asks for a password when opened, it has to be unlocked before its pages can be read.
• If you only have phone photos of the pages rather than a PDF, you can run a single image through Image OCR to get the text, or turn the photos into a PDF first.

Step by step: OCR a scanned PDF in the browser

1. Open OCR PDF and drag your scan onto the upload area. A preview of the pages appears.
2. In the language picker, select every language printed in the document and nothing else. English is preselected. For a form printed in English and Hindi, select both.
3. Leave Content type on Auto for normal letters and reports. Choose Dense paragraph / book for pages that are one solid block of text, or Sparse / scattered for forms where text sits in separate boxes.
4. Decide on Enhance image (on by default). The trade-offs are in the next section.
5. Press Make PDF Searchable. The first run downloads the language data; after that it is cached in the browser.
6. When the success message appears, check the Extracted Text tab for obvious misreads, then press Download Searchable PDF. Copy text and a .TXT download are there too if you only need the words.

Our test: how scan resolution changes accuracy

We wrote a one-page notice of 204 words, printed it to A4 page images at three resolutions with a slight tilt and speckle noise, saved each as a PDF, and ran each through the tool twice, with Enhance image on and off. We then pulled the text back out of each downloaded PDF and compared it word for word with the original.

Three things stand out. First, resolution is the biggest lever: nothing we could toggle rescued the 100 DPI page, where “rupees” became “mpees” and “librarian” became “libranan”. The Tesseract project's own quality guide says the engine works best at 300 DPI or more, and our results line up with that. Second, Enhance image did not help on these computer-made test pages, and it cost about four points of accuracy on the worst one. Third, the output file can be much larger than the input, because the tool stores a freshly rendered page image rather than reusing the original JPEG.

Treat these as indicative, not a benchmark. They come from one page, one typeface and synthetic noise, not from a real flatbed scanner, and a faded photocopy or a phone photo taken at an angle will behave differently.

Enhance image: when to leave it on

With Enhance image on, each rendered page is converted to grayscale and its contrast is raised before recognition. That same processed image becomes the visible page in your download. In our 300 DPI test the blue stamp on the page came out grey. With the toggle off, the stamp kept its colour, but the file was nearly five times larger.

• Leave it on for faint, low-contrast or yellowed paper, and whenever a smaller output file matters more than colour.
• Turn it off when colour carries meaning (signatures in blue ink, highlighted amounts, coloured stamps) or when a sharp, clean scan reads fine without help.
• If a low-resolution page gives a poor result, run it once with each setting and keep the better text. On our 100 DPI page the plain version won.

The order of steps that keeps your text layer

Our Compress PDF tool works by drawing every page again as a JPEG image. That is effective for shrinking scans, but it also flattens away any invisible text. If you compress a searchable PDF, the result looks the same and is no longer searchable. So when a portal needs a small file that you can also search, compress the scan first and run OCR on the compressed copy.

The last row is worth knowing about. The PDF editor runs the same OCR engine when you open an image-only file, and the recognised words become editable lines, so you can repair “moming” before saving instead of living with it.

Getting a better scan in the first place

1. Scan at 300 DPI in grayscale. Colour scanning triples the data without improving recognition of black text.
2. Lay the page flat and square on the glass. Tilt, curved pages near a book spine and shadows all cost accuracy.
3. For phone photos, shoot straight down in even daylight, fill the frame with the page, and avoid zooming in digitally.
4. Keep handwriting out of your expectations. The engine is trained on printed text, so handwritten notes in margins will be missed or turned into noise.

Frequently asked questions