# How to Make a PDF Searchable: Find the Cause First

- Author: Schema Law (https://www.schemalaw.com/)
- Published: 2023-12-27T06:46:58Z
- Updated: 2026-10-11T19:22:20Z
- Category: Searchable PDFs
- Canonical: https://www.schemalaw.com/blog/how-to-make-a-pdf-searchable

> How to make a PDF searchable when you do not know why search fails: five causes, a short diagnosis, and the fix that matches each one, OCR or otherwise.

To make a PDF searchable, first find out why it is not. If the pages are pictures, run OCR and save a copy with a text layer. If the pages hold text that your search cannot use, or the words sit in attached files, OCR is the wrong tool, and a different fix applies.

Running OCR on every stubborn file is tempting, and for a plain scan it is the answer. On a file whose text was fine it is wasted work, and it can leave you with a recognized copy that is worse than the one you started with. Diagnose first, then convert.

In the short check, search a middle page for the plaintiff, the defendant, a Bates number, and the word exhibit, then try to select that line. Compare it with the page image. A motion for summary judgment that selects on the first page, and an affirmative defense that fails where the exhibits begin, tells you the cause before you convert the whole file.

## Why is this particular PDF not searchable?

A PDF fails to search for one of five reasons: its pages are scans, its pages are pictures made by software, only some pages carry text, the text is stored in a form search cannot match, or the words live in attached files and not on the pages. Four call for recognition; the fifth does not.

### The five causes

1. **A scan.** Paper went through a scanner, copier or fax, and every page is one picture. No characters exist in the file.
2. **A picture made by software.** The document began as text on a computer and was turned into an image along the way: printed with an image setting, flattened, or passed through a system that returns pictures. It looks sharper than a scan, and it is just as empty.
3. **A mixed file.** Some pages carry text and others are pictures. A typed cover letter in front of scanned exhibits is one example. Search works, finds a little, and gives you false comfort.
4. **Text that search cannot match.** The page holds characters, and you can drag across them, but the file does not record which letter each shape stands for. A search for a word that is plainly on the page returns nothing.
5. **Words that are not on the pages.** The PDF is a wrapper. Its content sits in attached files or in the component files of a PDF Portfolio, and a plain search of the open pages never reaches them.

### What is not on the list

Two situations look like an unsearchable file and are something else. A permission setting can stop copying while search still works, so test search itself before you blame security. And a sound file can fail to search in one program and succeed in another. If a second PDF reader finds the word, the file is fine, and the post on [PDF search not working](https://www.schemalaw.com/blog/pdf-search-not-working) covers program-side failures.

Knowing which of the five you face tells you whether to convert, to extract, or to pick up the phone.

## A five-minute diagnosis before you run OCR on an unsearchable PDF

Diagnose an unsearchable PDF by testing the cheap causes first: search in a second program, try to select a word, look at more than one page, paste a selected line into a blank document, then open the attachments panel. Each test either names the cause or rules one out.

Work through the steps in order and stop at the first one that gives an answer.

1. **Search in a second PDF reader.** Open the file in a different program and search for a word you can see. A hit means the file has usable text and the first program was the problem.
2. **Try to select one word.** Drag across a word on the page. If the pointer draws a box or grabs the whole page, the page is a picture. The post on why you cannot copy text from a PDF describes [the selection and search test](https://www.schemalaw.com/blog/why-cant-i-copy-text-from-a-pdf#how-can-you-tell-whether-a-pdf-has-real-text) in full.
3. **Repeat on a middle page and the last page.** If page one selects and page thirty does not, the file is mixed.
4. **Zoom to 400% on a picture page.** Soft, speckled letters on a gray or tilted background point to a scanner. Crisp letters on a pure white, perfectly square page point to a picture made by software.
5. **Paste a selected line into a blank document.** If words select but the paste is boxes, symbols or wrong letters, the text is there and search cannot match it.
6. **Open the attachments panel.** If the pages are a cover sheet and the panel lists files, or the PDF opens as a list of documents, the content is in the components.

### Reading the results

| What the tests show | Cause |
| --- | --- |
| Second reader finds the word | Program-side problem; the file is sound |
| No page lets you select a word; letters are soft at high zoom | Scan |
| No page lets you select a word; letters are crisp at high zoom | Picture made by software |
| Some pages select and others do not | Mixed file |
| Words select; pasted text is nonsense | Text that search cannot match |
| Pages are few or blank; attachments panel lists files | Words not on the pages |

Write the result in the file note before you fix anything. "Pages 1 to 3 text, pages 4 to 61 image, scanned" is a sentence the next person can act on.

One more thing to check while the file is open: has someone already run OCR? Document properties list the program that produced the file. If that program is an OCR tool or a scanner's software and search still fails, you are no longer diagnosing a missing text layer. The post on [making a PDF searchable when OCR has already run](https://www.schemalaw.com/blog/how-to-make-a-pdf-ocr-searchable) picks up from there.

## Which fix matches which cause of an unsearchable PDF?

Each cause of an unsearchable PDF has its own fix. Scans and software-made pictures need OCR. A mixed file needs OCR on its picture pages only. Text that search cannot match needs the pages treated as pictures and recognized afresh. Attached or component files need extracting, then handling one by one.

| Cause | Fix | First thing to try | What to avoid |
| --- | --- | --- | --- |
| Scan | Run OCR, save a searchable copy | Convert a copy of the file as received | Rescanning a printout of the scan |
| Picture made by software | Get the text version; OCR if you cannot | Ask the sender for the file before it was flattened or printed | Assuming a sharp page will be read perfectly |
| Mixed file | OCR the picture pages, leave the text pages alone | A tool or setting that skips pages already holding text | Forcing recognition over good typed text |
| Text search cannot match | Recognize from the page picture | Run OCR with the existing text ignored or replaced | Trusting copy and paste from the old text |
| Words not on the pages | Extract the components and treat each by type | Save each attachment out as its own file | Running OCR on the cover sheet and calling it done |

### Notes on the OCR fixes

For a scan, the procedure is short and belongs to another post: the guide to [making a scanned PDF searchable](https://www.schemalaw.com/blog/how-to-make-a-scanned-pdf-searchable) has the steps and a test for the result. In outline, you keep the received file, upload a copy, let the pages be read, and export a searchable PDF under a new name. In Schema Law the upload takes a PDF, PNG or JPEG of up to 20 MB, as the [three steps on the homepage](https://www.schemalaw.com/#how-it-works) set out.

For a mixed file, how a tool treats pages that already hold text is the setting that matters. Tools differ, and some describe the choice in their documentation. OCRmyPDF, an open-source tool, [documents a skip mode](https://ocrmypdf.readthedocs.io/en/latest/advanced.html) in which "no image processing or OCR will be performed on pages that already have text", and a force mode in which "all pages will be rasterized to images". The first keeps good typed text. The second replaces it with a guess. Whatever you use, find out which behavior you are getting, then compare a typed page before and after.

For text that search cannot match, the same force-style behavior is what you want, because the existing characters are the problem. The OCRmyPDF page lists "fixing OCR text with a damaged character map (text is selectable but not searchable)" among the uses of its force mode. When the unmatched text is confined to a logo, a chart or a pasted screenshot on an otherwise typed page, see the post on [making PDF text searchable](https://www.schemalaw.com/blog/how-to-make-pdf-text-searchable).

## PDFs printed to an image, flattened or downloaded from a portal as pictures

A PDF can be all pictures without ever touching a scanner. Printing with an image option, flattening a file to lock its contents, and downloading from a system that renders pages as images all produce the same result: sharp, clean pages with no characters behind them.

### How it happens

- **Print settings.** Some print dialogs and PDF printer drivers offer to send each page as an image. The output looks identical on screen and holds no text.
- **Export settings.** An export option can swap text for a picture of it. Microsoft's instructions for [saving Office files as PDF](https://support.microsoft.com/en-us/office/collab-files/save-or-convert-to-pdf-or-xps-in-office-desktop-apps) list one for Word, and the post on converting Word to a searchable PDF quotes it under [whether a PDF saved from Word is already searchable](https://www.schemalaw.com/blog/convert-word-to-searchable-pdf#is-a-pdf-saved-from-word-already-searchable).
- **Flattening.** Tools that merge form entries, markup or redactions into the page may do it by turning the whole page into an image.
- **Portals and platforms.** A records portal, a signing platform or a fax service may hand back a rendered picture of each page whatever was uploaded.

### Why it matters for the fix

A picture made by software is kind input for OCR: straight, high contrast, no specks. The output is still a recognition result, so names, figures and section numbers need reading against the page.

The better fix is upstream. Somewhere a version with real text exists or existed. If you made the file, re-export it with the image option off. If a colleague flattened it, the copy from before that step is the one to use. If a portal produced it, see whether the portal offers the document in another form. The post on [why a PDF is no longer searchable](https://www.schemalaw.com/blog/pdf-no-longer-searchable) traces which step in a workflow strips the text.

### A worked example

A fictional example. Devlin Achterberg, a paralegal, receives a 12-page settlement statement from a title company's portal. Every page is crisp, and no word can be selected. The zoom test shows clean letter edges on a white ground, so this is a software-made picture. He emails the escrow officer for the PDF as generated. The file that comes back has the same 12 pages, is far smaller and searches on every page. No OCR was run, so no figure is a recognition result that needs reading against the page.

## PDF portfolios and attached files that a search passes over

A search of the open pages does not reach files attached to a PDF, and a PDF Portfolio is a container of separate files, not one long document. If the words you want are in a component or an attachment, the wrapper will seem unsearchable until you search or extract the parts.

### What Adobe's documentation says

Adobe describes a PDF Portfolio as "multiple files assembled into an integrated PDF unit" in which "the original files retain their individual identities". The same [overview of PDF Portfolios](https://helpx.adobe.com/acrobat/using/overview-pdf-portfolios.html) warns that if you open one "in a browser, email app (like Outlook), or another PDF viewer, you might see a 'Get Adobe Reader' message." A reader that shows only that message has nothing on its pages to find.

For Acrobat itself, the [page on searching PDFs](https://helpx.adobe.com/acrobat/using/searching-pdfs.html) puts both cases in the separate Search window, under its additional options. The "Look In" setting can restrict a search to "all of a currently open PDF Portfolio", and "Include Attachments" is described as an option that "Searches files that are attached to the current PDF or other attached PDFs (up to two levels deep)." For other programs, read their documentation or test with a word you know is in an attachment.

### What to do with a wrapper

1. Open the file in a program that shows its attachments or components, and list them. Note each one's name and type.
2. Save each component out as its own file, into a folder named for the wrapper.
3. Diagnose each extracted PDF on its own. Some will be text, some scans.
4. Convert the scans. An OCR tool that takes PDF, PNG and JPEG uploads will not take the wrapper's spreadsheets or emails, and those do not need OCR in any case.
5. Keep the wrapper as received, beside the extracted folder.

Count as you go. If the wrapper lists nine components, the folder holds nine files. An attachment left inside the wrapper is a document nobody searched.

## When is asking for a better file the right fix instead of OCR?

Ask for a better file whenever a version with real text exists and the person holding it can send it. That covers pictures made by software, files flattened by a colleague, text that will not match, and scans of documents that were printed from a computer. OCR is the fix when paper is the only original.

### Signs that a better file exists

- The page is perfectly straight and clean, which suggests it never was paper.
- The document is of a kind produced by a system: a statement, a report, a generated form.
- The sender is the author, or works with the author.
- The file name or properties show an office program or a business system as the origin.

### What to ask for

Be specific, because "a better copy" gets you the same file again. A request that works:

> Could you send the PDF as it was saved from the program that created it, not a printed or scanned copy? I need to be able to search the text.

If the document is a spreadsheet or a long table and you intend to work with the figures, ask for the original file format as well.

### When to stop asking

Requests cost time, and some senders cannot or will not help. Give the request a deadline that fits the matter. If nothing arrives, convert what you have and note in the file that the text is a recognition result. Whether another party is obliged to give you a searchable or native version is a question for the rules and orders in your matter, and nothing here answers it.

The request can also come the other way: a court portal, a client system or opposing counsel rejects your file as unreadable. That is a different job with its own checks, covered in the post on [making a PDF OCR readable](https://www.schemalaw.com/blog/make-pdf-ocr-readable).

If you have reached the end of the diagnosis and the answer is plain OCR on a plain scan, the hub post on [creating a searchable PDF](https://www.schemalaw.com/blog/create-searchable-pdf) sorts the routes by starting point.

## 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)