PDF stands for Portable Document Format, and its main promise is that a document looks the same everywhere. It says much less about what the document contains underneath. Two PDFs that look identical on screen can hold completely different things: one a page of real text, the other a photograph of that page.

That difference explains most search failures.

Three kinds of PDF#

Every PDF page falls into one of three types.

A born-digital page was created by software, such as a word processor exporting to PDF. It contains real characters in a text layer. Search works, copy and paste works, and screen readers can read it.

An image-only scan is a picture of paper. It came from a scanner, a copier or a phone app, and it contains no text at all. Search finds nothing, because there is nothing to find.

A searchable scan is a picture of paper with an invisible text layer placed behind it by OCR software. The page looks like the scan, and search works through the hidden text. How well it works depends entirely on how accurate that OCR was.

A single file can mix all three. A contract might have typed pages exported from Word, a scanned signature page and a faxed exhibit, each behaving differently when you search.

How to tell which one you have#

Try these checks on a page that isn’t working:

  1. Select a word with your cursor. If you can highlight individual words, there is a text layer.
  2. Copy a sentence and paste it into a plain text editor. If it matches the page, the text layer is good. If it comes out garbled or missing, the text layer is poor or broken.
  3. Use the viewer’s search for a word you can see on that page. If it isn’t found, either there is no text layer or the word was misread.
  4. Open the document properties and look at the fonts list. Image-only scans usually list no fonts at all.

Why search fails on scanned pages#

There is no text layer#

The most common reason. Many scanners and copiers save image-only PDFs by default, and phone scanning apps often do too unless OCR is turned on. Until the page goes through OCR, no search will ever find anything on it.

OCR misread the word#

OCR reads characters by their shapes, and some shapes are easy to confuse. These mix-ups break search because a single wrong character means the word doesn’t match:

On the page Often read as
m rn
l (lowercase L) 1 or I
0 (zero) O
5 S
cl d
fi, fl a single unknown character
é, ọ, ụ e, o, u, or a stray mark

Names and numbers suffer most, because the software can’t check them against a dictionary. A search for a client named “Mahmoud” fails if OCR produced “Mahrnoud.”

The word was split across lines#

Typeset documents hyphenate long words at the end of a line. The text layer then holds “indem-” on one line and “nify” on the next, and a search for “indemnify” misses it. Some OCR and cleanup tools rejoin hyphenated words, and many don’t.

Headers, footers and page numbers get in the way#

When a sentence runs from one page to the next, the text layer often holds the footer, the page number and the next page’s header in the middle of it. A search for a phrase that crosses a page break fails, and an exported text file reads like “the parties agree Page 4 of 12 CONFIDENTIAL that the”.

The reading order is wrong#

On pages with columns, sidebars or tables, OCR has to decide what order to read things in. If it reads straight across two columns, words land next to strangers and phrase searches fail. Tables are especially hard, because each cell may end up on its own line or merged with its neighbors.

The text is there but encoded strangely#

Some born-digital PDFs have a text layer that can’t be read properly. The page displays correctly, but the fonts lack a map from their internal codes to real letters, so copying produces gibberish and searching finds nothing. Older exports, some legal and accounting software and some PDF printers produce files like this. Running OCR on the page usually fixes it, because OCR ignores the broken text and reads the image.

The scan was too poor to read#

Low resolution, a crooked page, faint photocopies, stains and shadows all lower OCR accuracy. If the scan was made at 150 pixels per inch, a better OCR engine won’t recover the missing detail. Re-scanning at 300 pixels per inch or more is the only real fix.

The language wasn’t set#

OCR engines use language data to decide what’s likely. A page in Igbo, Yoruba, French or Spanish read with English-only settings loses accents and special letters, and searches that include them fail.

Your search tool only indexes some files#

Document management systems, email archives and desktop search tools may skip large files, image-only PDFs or files above a page limit when they build their indexes. A file that is searchable when opened can still be invisible to the firm’s search.

How to make scanned PDFs searchable#

Run OCR on everything that needs it#

Start by OCR-ing image-only pages, and re-run OCR on pages with garbled text layers. Keep the original image and add the text behind it, so the page still looks like the scan.

Clean up what OCR produced#

Better results come from removing the noise around the text. That means stripping repeated headers, footers and page numbers, rejoining broken lines and hyphenated words, and keeping tables as tables. These steps matter most when you export to text or Word, or feed the content into a search index or AI tool.

Improve the source when it matters#

For important records with poor scans, re-scanning at a higher resolution, in grayscale and with the pages flattened, usually beats any amount of software correction.

Search smarter in the meantime#

While you work through a backlog, a few search habits help:

  • Search for part of a word, such as “indemn,” to catch hyphenation and endings.
  • Search for a second distinctive word in the same passage.
  • Use fuzzy or approximate search if your tool supports it.
  • For names, try likely misreadings, such as “rn” for “m.”

Test one page before converting a thousand

Before batch-processing an archive, run OCR on ten representative pages and search them for names, numbers and dates you can see. If those come back reliably, the settings are good enough for the rest.

Stop the problem at the source#

Fixing old files is slower than creating good ones. A few changes stop new unsearchable PDFs from piling up:

  • Set every office scanner and copier profile to save a searchable PDF, with OCR turned on and the right language selected.
  • Set the default resolution to at least 300 pixels per inch.
  • Turn on OCR in phone scanning apps that staff use for receipts and signed pages.
  • Ask clients and vendors to send born-digital PDFs where they can, instead of printing and scanning.
  • When you combine files into one PDF, check that the scanned pages in the bundle went through OCR.

A short note in your office procedures, and a check of each scanner’s settings once a year, prevents most of the trouble described above.

How Sabi PDF helps#

Sabi PDF reads scanned pages automatically and keeps the real text on pages that already have it. You can also tell it to re-read every page, which fixes files with garbled text layers. It turns sideways scans upright, removes repeated headers, footers and page numbers, fixes broken lines and keeps tables as tables in Word and PDF output. Its searchable scan export keeps your original pages and makes the text on scanned ones selectable. It handles scans, phone photos saved as PDF, plan sets and messy exports up to 25 MB, and it deletes files automatically after 60 minutes.

Questions people ask

How can I tell whether a PDF has searchable text?

Try to select a word with your cursor. If you can highlight individual words and copy them into another document, the page has a text layer. If the selection draws a box around the whole page or nothing happens, the page is an image with no text.

Why does copied text from my PDF come out as gibberish?

The PDF is using a font without a proper map from its internal codes to real characters. The page displays correctly, but the text underneath is a string of codes your computer can't translate. Running OCR on that page usually produces usable text.

Why does search find a word on some pages and not others?

Usually because OCR misread it on the pages where it fails, or because the word was split by a hyphen at the end of a line. Try searching for part of the word, or a different distinctive word nearby.

Will making a PDF searchable change how it looks?

Not if you use a searchable scan format. It keeps the original page image and adds an invisible text layer behind it, so the page looks the same.

Sources

  1. Tesseract OCR, project README (GitHub)
  2. Sabi PDF, cleanup and export features

Keep reading

This guide is general information for small firms. It is not legal, insurance or compliance advice for your situation.