# PDF No Longer Searchable? What Stripped the Text Layer

- Author: Schema Law (https://www.schemalaw.com/)
- Published: 2022-10-14T11:44:44Z
- Updated: 2026-10-11T19:22:20Z
- Category: Searchable PDFs
- Canonical: https://www.schemalaw.com/blog/pdf-no-longer-searchable

> A PDF is no longer searchable when a later step rebuilt its pages as pictures. See which steps do it, how to find the one in your workflow, and the fix.

A PDF that is no longer searchable was rebuilt somewhere along the way. A step such as printing to PDF, flattening, a size reduction, a signing platform or a fax made new pages from the look of the old ones and left the text behind. Find the last copy that still searches, and the step after it is your cause.

Nothing is wrong with the reader, and the text did not decay. Text in a PDF does not wear out with time or with copying. If the file searched on Monday and does not on Thursday, a program wrote a different file in between. This guide lists the programs that do it, shows how to find the one in your own workflow, and sets out the two repairs.

On a version that fails, search the motion title and the plaintiff and defendant under the judge's name. Then search dismissal or order on a later page. Try to select the same line and set it beside the page image. If those words selected earlier in the week and the picture still shows them, a step in between rebuilt the pages and left the text behind.

## What makes a PDF stop being searchable?

A PDF stops being searchable when a program writes a new version of it from the appearance of the pages and not from their contents. The picture of each page is carried over because it is what you see. The text is dropped because, in a scanned and recognized file, it is invisible by design.

It helps to hold two kinds of operation apart.

**Operations that copy pages.** Saving under a new name, moving the file, attaching it to an email, zipping it, extracting a page range, and merging whole files in a PDF editor all pass the page contents along as they are. The text travels with the page.

**Operations that redraw pages.** Printing, flattening, converting to an image format and back, some kinds of optimization, and any system that renders the document in order to show or transmit it. These produce pages that look the same and are built differently.

A searchable scan is fragile under the second kind in a way a typed PDF is not. In a file saved from a word processor, the text is the visible content, so a program that redraws the page still has letters to draw. In a searchable scan, the visible content is a picture and the text sits out of sight. A program that asks "what does this page look like?" gets the picture and nothing else. The post on [how to make a scanned PDF searchable](https://www.schemalaw.com/blog/how-to-make-a-scanned-pdf-searchable#what-makes-a-scanned-pdf-searchable) explains how that hidden text is stored.

### Signs that pages were redrawn

| Sign | What it suggests |
| --- | --- |
| No word can be selected on any page | Every page was rebuilt as a picture |
| Typed pages still select, scanned pages do not | Only the invisible text was dropped |
| The file's creator or producer property names a print driver or a platform | That program wrote the current file |
| Bookmarks, links and comments have gone too | The whole document was regenerated, not edited |
| The file is much smaller or much larger than before | The pictures were recompressed or re-rendered |

Open the document properties in your reader and read the producer field first. It names the software that last wrote the file, and it is the fastest single clue you have.

## Printing to PDF, flattening and saving as an image: the usual culprits

Three actions on your own computer can strip the text from a searchable PDF: sending it to a PDF printer, flattening it, and saving it as images. Each is a reasonable thing to do for another purpose, and each can build new pages from what the old ones look like. Some size-reduction settings are a fourth.

### Printing to PDF

A print command sends a program's drawing of the page to a printer driver. A PDF printer is a driver that writes that drawing into a new file. Text that is not drawn on the page, which is how the recognized text of a scan is stored, may not be in the drawing at all. Whether it survives depends on the program doing the printing and the driver receiving it, so do not assume either way. Print one page of a searchable scan to PDF with your own setup and try to select a word.

People print to PDF for sensible reasons: to pull pages 12 to 15 out of a long file, to combine a document with its comments, to remove form fields, or because Print is the menu command they know. For pulling pages, use the editor's extract or split command, which copies pages and keeps their text. The post on how to [split a searchable PDF](https://www.schemalaw.com/blog/split-a-searchable-pdf) compares the two methods.

### Flattening

"Flatten" means different things in different programs. In some it merges comments and form fields into the page and leaves text alone. In others it turns every page into one picture. The second kind removes the text layer along with everything else that was separate from the image. If you flatten deliberately to remove hidden content, that result is the point; the section on [what flattening a PDF to images removes and what it keeps](https://www.schemalaw.com/blog/convert-pdf-to-non-searchable-pdf#what-flattening-a-pdf-to-images-removes-and-what-it-keeps) covers that use. If you flattened only to lock down annotations, read your program's documentation to learn which kind of flattening it performs.

### Saving as images

Exporting a PDF to JPEG, PNG or TIFF and rebuilding a PDF from those images discards the text for certain, because an image file has nowhere to keep it. The same happens when a page is captured with a screenshot tool and pasted into a new document, or when a PDF is dragged into a program that imports each page as a picture.

### Size reduction

A "reduce file size" or "optimize" command can be harmless or destructive depending on its settings. Recompressing the pictures leaves the text alone. A setting that discards hidden content, or that rasterizes pages, does not. The section on [which size-reduction settings quietly remove or damage the text layer](https://www.schemalaw.com/blog/searchable-pdf-file-size#which-size-reduction-settings-quietly-remove-or-damage-the-text-layer) goes through the settings to watch.

| Action | What it does to pages | Text layer afterwards |
| --- | --- | --- |
| Save As, copy, rename, email | Copies them | Kept |
| Extract or split pages in a PDF editor | Copies them | Kept |
| Print to PDF | Redraws them | Test with your setup |
| Flatten annotations only | Edits them | Kept, per the program's documentation |
| Flatten to image | Replaces them with pictures | Gone |
| Export to images and rebuild | Replaces them with pictures | Gone |
| Optimize with "discard hidden content" or similar | Edits them | At risk |

## Does redacting, stamping or numbering a PDF remove its text layer?

Redacting, stamping and numbering are meant to change only part of each page, and a tool that works that way keeps the remaining text. Some tools take a blunter route and rebuild every page as a picture with the marks burned in. Which one you have is a fact about your software, so test it on a searchable sample.

### Redaction

A redaction tool that works correctly removes the text under the redacted area along with the image of it, and leaves the text on the rest of the page searchable. A tool or setting that converts each redacted page to an image removes all the text on those pages. Neither result is a fault; they are two designs, and the second is sometimes chosen on purpose so that nothing can be left behind.

The consequence for search is that a redacted production may need recognition run again before the remaining text can be found. The order in which you recognize and redact, and how to test that no text survives under a box, belong to the post on [whether to OCR before or after redaction](https://www.schemalaw.com/blog/ocr-before-or-after-redaction).

### Stamps and exhibit stickers

A stamp added as an annotation sits on top of the page and changes nothing beneath it. The text is lost only when the stamp is then merged into the page by a flatten-to-image step or by printing to PDF. If your process is "stamp, then print to PDF so the stamp cannot be moved", the print is what strips the text, not the stamp.

### Bates and page numbering

Numbering adds a short piece of text or a small image to each page. Done by a PDF editor's numbering command on the existing file, it leaves the page's text in place. Done by a tool that renders each page to an image in order to draw the number onto it, it does not. Vendors and litigation-support tools differ, and so do the options within one tool.

A numbered scan that was never searchable to begin with can also mislead you. The Bates number selects, so the file passes a quick "can I select text?" test, while the body of every page is still a picture.

### A sample test for any marking tool

1. Take a five-page searchable PDF that you have confirmed finds a word on every page.
2. Apply the redaction, stamp or number with your normal settings and save as you normally would.
3. Search the result for a word on a marked page, away from the mark.
4. Search for a word on an unmarked page.
5. Record the tool, version, settings and result in a note the whole team can find.

Run the test once for each tool and again after a software update. It tells you whether that step belongs before or after recognition in your process.

## E-signature platforms, upload portals and fax: systems that hand back a picture

Systems outside your computer can return a document that looks like the one you sent and carries none of its text. A fax transmits pictures of pages by its nature. Signing platforms, client portals, court and agency filing systems, and messaging apps may each reprocess an upload, and their output is what you get back.

### Fax

A fax is an image transmission. Whatever you send, the receiving end gets a picture of each page, at fax resolution. A searchable PDF sent to a fax number arrives as an unsearchable one, and a document faxed to you and saved as a PDF has no text until someone runs recognition on it. The post on how to [convert a fax to a searchable PDF](https://www.schemalaw.com/blog/convert-fax-to-searchable-pdf) covers the received side.

### E-signature platforms

A signing platform takes your upload, places signature fields on it, and returns a completed document when everyone has signed. What that completed file contains is determined by the platform. This post makes no claim about any named service. Test yours: upload a searchable scan, sign it yourself as a test, download the completed document, and try to select a word in the body. Then read the platform's documentation for what it says about the completed file.

A related case arises when a signed document comes back on paper. A client prints the agreement, signs it in ink, scans it on a home printer and emails it. The file you now hold is a new scan with no connection to the searchable PDF you sent.

### Upload and filing portals

A portal can store exactly what you upload, or it can convert the upload into its own format, add a header or a stamp to each page, combine it with a cover sheet, or compress it. The version you download later may be the converted one. Courts and agencies publish their own technical requirements, and they vary; the post on [searchable PDFs for court filing](https://www.schemalaw.com/blog/searchable-pdf-for-court-filing) deals with what courts ask for. For any portal, the practical step is to download what you uploaded and compare.

### Email, chat and phone apps

Sending a PDF as an email attachment does not alter it. Other routes can. A messaging app that shows a preview may send a picture of the first page in place of the file. A phone's share sheet may offer to send a document "as photo". A scan-to-email feature on a copier produces a new scan each time. When a file reaches you in an unexpected condition, ask how it was sent before assuming how it was made.

| Channel | What may happen to the file | How to find out |
| --- | --- | --- |
| Fax | Always becomes page pictures | No test needed |
| Signing platform | Completed document is generated by the platform | Sign a test document and inspect it |
| Client or vendor portal | Upload may be converted or stamped | Download your own upload and compare |
| Court or agency system | Varies by system | Read its technical requirements; compare a download |
| Messaging app | May send a preview image | Ask the sender for the file itself |

## How do you find the step in your workflow that stripped the text?

Work backwards through the saved versions of the file until you reach one that searches. The step between that version and the next is the one that stripped the text. If no earlier versions were kept, rebuild the workflow on a test file and check after every step.

### When you have earlier copies

1. **List every version you can find.** Look in the matter folder, the sent and received email, the downloads folder, the document system's version history, and any portal the file passed through.
2. **Sort them by modified time.**
3. **Test each for text, oldest first.** Open the file, go to a page in the middle, and try to select one word. Then search for a word you can see on that page.
4. **Stop at the first version that fails.** Read its producer property and note it.
5. **Name the step.** Ask what was done between the last good version and the first bad one, and by which program.

### When you have no earlier copies

Rerun the process on a short searchable sample and test after each action.

A worked example, with invented names. A firm prepares an exhibit set for *Marchetti v. Oster Holdings*. The process on paper is: scan, recognize text, combine, number, add exhibit stickers, reduce size, upload to the client portal. The paralegal runs a ten-page sample through it and records one line per step.

| Step | Program | Selectable after? | Search finds "invoice" on p. 6? |
| --- | --- | --- | --- |
| Scan | Copier | No | No |
| Recognize text | OCR tool | Yes | Yes |
| Combine | PDF editor | Yes | Yes |
| Number | PDF editor | Yes | Yes |
| Exhibit stickers, then "print to PDF to lock" | PDF editor and print driver | No | No |
| Reduce size | PDF editor | No | No |
| Upload and download | Client portal | No | No |

The text disappears at the fifth step, and it is the printing, not the stickers, that removes it. The two later steps look guilty only because they come afterwards. The fix is to drop the print step or to move recognition after it.

### Writing the result into the process

Once you know the step, change the written process, not only today's file. Either replace the step with one that copies pages, or move recognition so it runs after the last step that redraws them. Then add one line to the end of the process: "search the final file for a word on a middle page before it leaves the office".

## Restoring search: return to the earlier copy or run OCR again

You have two repairs. Go back to the last version that still has its text and redo the later steps in a way that keeps it, or run recognition again on the stripped file. Prefer the earlier copy when it exists, because its text is the text you already searched and quoted, and its pictures went through fewer conversions.

| | Return to the earlier copy | Run OCR again on the stripped file |
| --- | --- | --- |
| Needs | A saved version that still searches | Only the file you have |
| Page pictures | As they were before the stripping step | Whatever the stripping step left, which may be coarser |
| Text | The same text as before | Newly recognized, possibly different in places |
| Later marks (numbers, stamps, redactions) | Must be reapplied | Already on the page, and read as part of it |
| Best when | The stripping step can be replaced | The marks must stay exactly as they are |

### Returning to the earlier copy

Copy the last good version, then redo each later step using a method that copies pages: extract in place of print, annotation flattening in place of image flattening, an optimizer setting that keeps hidden text. Test after each step this time.

### Running OCR again

When the stripped file is the only one, or its marks cannot be redone, recognize it again. Three points apply.

- **Recognition reads what the picture now shows.** If the stripping step lowered the resolution, results on small print may be worse than the first time. Bates numbers and stamps that were burned into the page will be read as text, which is useful for searching by number.
- **Redacted pages need care.** Running recognition on a properly redacted, flattened page is safe because nothing lies under the boxes. Never go back to the unredacted version to recover text for a document that is going out.
- **The new text is not the old text.** If anyone quoted from or cited the earlier searchable copy, check those passages against the new one.

Any tool that takes a PDF and returns a searchable PDF will do this job. In Schema Law you upload a PDF of up to 20 MB, the pages are read, and Searchable PDF is one of the export formats. Every plan includes it, and the Free plan converts one document a day of up to three pages. The [formats section of the homepage](https://www.schemalaw.com/#formats) describes that export as your original file with a hidden text layer.

### Keeping it from happening again

- Keep the recognized file as its own saved version before any marking or sending step.
- Make "search a middle page" the last action before a file leaves the office.
- When a file comes back from outside, test it on arrival and note the result in the file name, such as `_ocr` for one that searches.

## 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)