# Split a Searchable PDF and Keep Every Part Searchable

- Author: Schema Law (https://www.schemalaw.com/)
- Published: 2026-06-29T15:58:41Z
- Updated: 2026-10-11T19:22:20Z
- Category: Searchable PDFs
- Canonical: https://www.schemalaw.com/blog/split-a-searchable-pdf

> Split a searchable PDF by extracting pages and each part keeps its text layer. Where to cut, how to name the parts, and how to prove no page was dropped.

You can split a searchable PDF without losing the text layer as long as you extract the pages into new files. Extraction copies each page whole, hidden text included. Printing a page range to a new PDF is a different operation, and it can leave you holding parts that are pictures only.

## Does splitting a searchable PDF keep the text in each part?

Yes, when the pages are extracted and not reprinted. The hidden text of a searchable PDF is stored page by page, alongside each page's picture. A split or extract command lifts whole pages into a new file, so each part arrives with the same text its pages had in the long file.

Nothing is recognized again during a split, so the parts are neither better nor worse than the source. A page that searched badly in the long file searches badly in its part. A page that had no text layer, such as a scanned exhibit merged in later, still has none.

Open every part and search for the pleading or the exhibit that belongs in it, plus a Bates number from its first page and its last. Compare the hit with the page image. A cross-claim or a deposition that starts in one part and finishes in the next will miss a phrase that crossed the cut. Move the break to the document boundary and split again.

What a split can change is everything that belonged to the file and not to a page: bookmarks, the document title, and the page labels a reader shows in its counter. Check those in each part if you depend on them.

To pull out only the pages that mention a particular term, the method is different; see the post on [extracting pages containing a word from a PDF](https://www.schemalaw.com/blog/extract-pages-containing-a-word-from-a-pdf).

## Where should a long searchable PDF be cut: document breaks, not page counts

Cut where one document ends and the next begins. A part that holds whole documents can be read, cited, sent and searched alone. A part that stops at page 250 because 250 was the limit leaves a contract in one file and its signature page in another.

A split by fixed page count is quick to set up and hard on the reader. A phrase that ran from the foot of page 250 to the top of page 251 can no longer be found by one search, and whoever opens part two starts in the middle of a sentence.

Find the breaks before you cut:

1. Open the bookmarks panel. If the file has bookmarks, they mark the documents.
2. If it has none, scroll the thumbnails and note each page that starts something new: a caption, a letterhead, an exhibit cover sheet.
3. Write the breaks as a list of page ranges.
4. Group neighbouring ranges until each group fits your limit.

A page or size limit is one reason to split. Schema Law caps one document at 3 pages on the free plan, 250 on Solo and 1,000 on Team, and takes files up to 20 MB; the [pricing page](https://www.schemalaw.com/pricing) has the figures. Suppose a 540-page records set in the fictional matter of Lindqvist v. Arbor Mutual must go in under a 250-page cap. The breaks fall after pages 198 and 421, giving parts of 198, 223 and 119 pages. No part is full, and no record is cut in two.

When a single document is longer than the limit, cut it at a section heading and say so in the part's name. Whether a file should hold one document or many in the first place is covered in the section on [one PDF per document or one per scanning session](https://www.schemalaw.com/blog/scanned-images-to-pdf#one-pdf-per-document-or-one-per-scanning-session).

## Extracting pages versus printing a page range to a new PDF

Extract. An extract or split command copies the chosen pages, with their text, into a new PDF. Printing a page range to a PDF printer sends the pages through a print driver, which draws them again, and what the driver writes may be a picture of each page with no text behind it.

The two routes can sit side by side in a program's menus, and they give files that look the same on screen.

| | Extract or split pages | Print a range to PDF |
| --- | --- | --- |
| What it does | Copies whole pages | Draws the pages again through a driver |
| Hidden text | Kept | Kept or lost, depending on the driver and settings |
| Page picture | Unchanged | May be resampled or recompressed |
| Bookmarks and links | May be kept for the extracted range | Lost |
| How to tell afterwards | Search works in the part | Search finds nothing, or you test and it happens to work |

If printing to PDF is the only tool on the machine in front of you, split one short range and test it before you do the rest. Search for a word on the first page of the part and on its last. No result means the text is gone, and the route has to change.

Check one more thing in the extract dialog: an option to delete the pages from the source after extracting. Leave it off. The long file is your record of what the set looked like before it was cut. A file that lost its text during some other step is the subject of the post on [a PDF that is no longer searchable](https://www.schemalaw.com/blog/pdf-no-longer-searchable).

## Naming the parts of a split searchable PDF so they sort and rejoin

Give every part the same base name, a part number with a leading zero, and the page range it came from. A folder then lists the parts in page order, anyone can see which pages a part holds without opening it, and a missing part is obvious from the gap in the numbers.

A pattern that works:

```
Lindqvist_Records_part01_p001-198.pdf
Lindqvist_Records_part02_p199-421.pdf
Lindqvist_Records_part03_p422-540.pdf
```

Each element does a job.

- **Base name.** Identical across parts, so they sit together in the folder.
- **`part01`, not `part1`.** Without the leading zero, a folder puts `part10` ahead of `part2`.
- **Page range of the source.** Three digits when the source has three-digit page numbers. The ranges should run on from one another with no gap and no overlap.
- **A suffix for status.** Add `_ocr` only to parts that have been through recognition, so a searchable part is never confused with an unrecognized one.

Avoid names that describe content in place of position, such as `Lindqvist_radiology.pdf`. They read well and cannot be put back in order. Put the description in the bookmark or in an index, and keep the position in the name.

When the parts need to become one file again, the names carry the order. The procedure is in the section on [putting the parts of a split scan back together in page order](https://www.schemalaw.com/blog/combine-searchable-pdfs#putting-the-parts-of-a-split-scan-back-together-in-page-order).

## How do you confirm no page was dropped when a searchable PDF was split?

Add up the page counts of the parts and compare the total with the page count of the source. Then open each part and confirm that its first and last pages are the pages its file name says they are. The sum catches a missing or repeated range; the boundary check catches a cut made one page off.

Record the source page count before you split anything. Then fill in a reconciliation like this one:

| Part | Range in name | Pages expected | Pages in file | First page checked | Last page checked | Search test |
| --- | --- | --- | --- | --- | --- | --- |
| 01 | 1 to 198 | 198 | 198 | Intake form | Discharge summary, final page | "Lindqvist" found |
| 02 | 199 to 421 | 223 | 223 | Radiology cover sheet | Billing ledger, final page | "Lindqvist" found |
| 03 | 422 to 540 | 119 | 119 | Adjuster's letter | Certification of records | "Lindqvist" found |
| Total | | 540 | 540 | | | |

Work out "pages expected" from the range: last page minus first page, plus one. Leave out the plus one and a column of ranges appears to be short by one page per part.

The total proves less than it seems to. A part that lost page 300 and gained a duplicate of page 301 still has the right count. That is what the boundary columns are for, and for a set that matters, add a third check: compare the last page of each part with the first page of the next, side by side with the source, to see that they are consecutive.

Finish with the search column. Search each part for one word you know is in it. A part with the right pages and no hits was produced by a route that dropped the text, and needs to be extracted again from the source.

Keep the reconciliation with the parts. If the split copies will be searched in place of the source, the homepage sets out [how a document is uploaded and exported](https://www.schemalaw.com/#how-it-works) for any part that still needs recognition.

## Try it on a real document

Upload a scanned PDF and export it as Word, a searchable PDF or plain text on the free plan. Your documents are never used to train an AI model. [Start free](https://www.schemalaw.com/api/auth/google)