# Make Scanned PDF Searchable Before You Convert It

- Author: FastBankConvert (https://www.fastbankconvert.com)
- Published: 2026-09-12
- Updated: 2026-09-12
- Category: Scanned and image statements
- Canonical: https://www.fastbankconvert.com/blog/make-scanned-pdf-searchable

> How to make a scanned PDF searchable with an OCR text layer, check that the layer lines up with the page, and what it changes for conversion.
To make a scanned PDF searchable, run it through OCR in a tool that saves a searchable PDF. The tool keeps each page picture and places an invisible layer of recognized text behind it. Then test that the layer lines up with the page and that a search for an amount finds it, before you trust the file for conversion.

## What does a searchable scanned PDF contain?

A searchable scanned PDF contains two things on every page: the original scan as a picture, and a hidden text layer that OCR placed exactly behind the words it recognized. You still see the picture. Search, selection and copying use the hidden text. Nothing about the visible page changes.

That design has a consequence people often miss. The page looks exactly as it did before, so the file seems unchanged and trustworthy. Every search hit and every copied figure, though, comes from OCR's guesses, not from the picture you are looking at. A 7 you can see clearly on screen might sit over a hidden 1.

The layer also has positions, not only characters. Each recognized word is placed at a spot on the page that should match the word in the picture. Good tools get the positions right to within a fraction of a character. Poor ones, or a skewed page, can leave the hidden word several millimetres away from the visible one. Positions matter because converters use them to build columns.

Files made this way are sometimes called "PDF with hidden text", "sandwich PDF" or "image over text", depending on the software. All of them mean the same thing.

## Tools that make a scanned PDF searchable

Three kinds of tool add the text layer: scanner software that saves directly as a searchable PDF, PDF editors with a recognize-text command, and dedicated OCR programs, including free command-line ones such as the open-source OCRmyPDF. They all follow the same pattern, and the settings that matter are much the same in each.

The settings to check before you run it:

- **Language.** Pick the language the document is written in. The wrong language lowers accuracy on names and on accented characters.
- **Pages.** Choose all pages, or only image pages if the file mixes scanned and digital pages and the tool can leave existing text alone.
- **Output type.** Choose the searchable option that keeps the original image. Avoid any option that replaces the page with redrawn text, covered below.
- **Deskew and cleanup.** If offered, straighten pages before recognition. A level page gives better text and better positions.
- **Resolution handling.** Leave downsampling off, or set it no lower than the scan's own resolution.

Scanning to searchable PDF at the source is the least effort when you have the paper in hand. Many scanners and copiers offer it as a save format, and it saves a second pass later.

The scanning itself has to be in order first. A statement printed on both sides that comes off a single-sided scanner in the wrong sequence will be just as jumbled after OCR. [Scanning a double-sided bank statement in order](/blog/scan-double-sided-bank-statement) covers that step.

## Checking that a searchable PDF text layer lines up with the page

Check the text layer with a few quick tests: select a line and see where the highlight falls, search for a word and an amount you can see, copy a row into a text editor, and add up a column against the printed total. Together they show whether the hidden text is present, placed correctly and accurate.

The six tests, in the order to run them:

1. **Select.** Drag across a line. If nothing highlights, there is no layer on that page.
2. **Highlight position.** Look at where the highlight sits. It should cover the visible words closely. A highlight that floats above, below or to the side means the layer is misaligned.
3. **Search a word.** Search for a word printed on the page, such as "Balance". Every page where it appears should report a hit.
4. **Search an amount.** Search for one specific figure, such as 1,284.50. Amounts are where OCR errors matter most, and a failed search here is the clearest early warning.
5. **Copy and paste a row.** Copy one transaction line into a plain text editor. Check that the date, description and amount come out in the right order with the right characters.
6. **Compare a total.** Copy a short column of amounts into a spreadsheet, sum it and compare with the page's printed total.

### What a misaligned layer looks like

Here is an example. On a crooked scan of a statement page, you select the line "09/14 CITY WATER 62.10". The highlight covers the date and description, but at the right of the line it drifts up and covers part of the row above. Searching for 62.10 finds it, so the characters are right. Copying the row, though, gives "09/14 CITY WATER 118.00", taking the amount from the line above. A converter reading positions would pair the wrong figures in exactly the same way.

The fix for that case is to straighten the page and run OCR again, not to edit the text. When test 4 fails because a digit was misread rather than misplaced, [the digit errors OCR makes on bank statements](/blog/bank-statement-ocr-digit-errors) lists the usual suspects. Dates have their own failure patterns, such as day and month swapping, which [OCR misreads of dates on bank statements](/blog/ocr-misread-dates-on-bank-statements) explains.

## Does a searchable PDF convert better to Excel?

A searchable PDF converts to Excel through routes that need text, such as Excel's own PDF import or a copy and paste, which see nothing in an image-only scan. The quality of the result is the quality of the text layer. Adding the layer gives those routes something to read. It does not make the underlying recognition any better.

What changes once the layer exists:

- **Text-only tools start working.** Importers that read PDF text now find characters and positions instead of an empty page.
- **Errors are baked in.** Any misread digit is now stored as text, and converters that trust existing text will pass it through without a second look.
- **A better OCR may be bypassed.** Some converters skip their own recognition when a page already has text. If your layer came from a weak tool, a converter with stronger OCR may never get the chance to improve on it.

That last point is worth testing. Convert the searchable file and the original image-only scan with the same converter, and compare the totals. If the original scan does better, keep using it as the input and use the searchable copy only for finding things.

When a converter rejects the file outright, the text layer is only one of several possible causes. [Finding the cause when a PDF to Excel conversion fails](/blog/pdf-to-excel-conversion-failed) runs through the others in order.

## Searchable PDF output options that alter the page image

Some save options change the page picture as well as adding text. The two to watch are image downsampling, which lowers the scan's resolution to shrink the file, and output modes that replace the scanned image with redrawn text. Either can leave you with a searchable file that has lost detail you may need later.

The options, and what they do:

| Option (names vary by tool) | What it does | Risk |
| --- | --- | --- |
| Searchable image, keep original | Adds hidden text, leaves the picture alone | None to the image |
| Searchable image, compact | Adds text and compresses or downsamples the picture | Small print can soften, harming any later OCR |
| Editable text and images | Replaces the scan with recognized text in a matching font | A misread is now what you see, with no picture to check against |
| Black and white conversion | Turns the scan into pure black and white | Faint figures and shaded rows can vanish |

Keep the original scan alongside any searchable copy. If the text layer turns out to be poor, you can run a better OCR tool on the untouched original, which is impossible once the picture has been compressed or replaced.

## When is making a scanned PDF searchable wasted effort?

Making a scanned PDF searchable is wasted effort when your converter does its own OCR, when the PDF already has real text, or when you only need the data once. In each case the extra step adds time and a second chance for errors, without giving the conversion anything it lacks.

Skip the step when:

- **The file already has text.** Try selecting a word first. Running OCR over a text PDF at best changes nothing and at worst replaces exact characters with guesses.
- **Your converter reads scans itself.** A tool built to handle scanned pages does the recognition as part of the conversion.
- **You will not search the file again.** If the goal is a spreadsheet and the PDF will be archived, the searchable copy has no later use.

Do it when the PDF is a record you will come back to, such as an archive of old statements you need to search by payee, or when the only conversion route you have reads text and nothing else.

FastBankConvert is built to read scanned statement PDFs as well as digital ones, though scanned pages take longer, and to tell you before download when the balances do not add up. It is getting ready to open, and [Get started](/get-started) puts you on the list.