# OCR Text Scanner: Does the Scanner or Software Read It?

- Author: Schema Law (https://www.schemalaw.com/)
- Published: 2024-05-22T16:08:04Z
- Updated: 2026-10-11T19:22:20Z
- Category: OCR basics
- Canonical: https://www.schemalaw.com/blog/ocr-text-scanner

> An OCR text scanner photographs the page; software does the reading. Where that software runs, what a copier's searchable PDF button does, and what to pick.

An OCR text scanner is a scanner sold with recognition software. The scanning hardware only photographs the page. The reading is done by OCR software, which may run inside the device, on the computer attached to it, or in a cloud service. Where it runs decides how good the text is and whether you can redo it.

That split is worth knowing before you buy equipment or set a firm-wide default. People shopping for the best OCR scanner are really choosing two things at once: a machine that takes good pictures of paper, and a program that reads those pictures. The two can be chosen separately, and often should be.

## Does an OCR text scanner read the page itself?

No. The scanner's glass and sensor capture an image, and nothing in that step knows a letter from a coffee stain. Recognition is a separate program. On some machines the program is built into the device, which makes it look as if the scanner reads. On others it runs on your computer or on a remote server.

There are three places the reading can happen, and a single office often has all three without anyone having chosen them.

| Where recognition runs | What it looks like to the user | Typical equipment |
| --- | --- | --- |
| Inside the device | You pick "searchable PDF" on the touch panel and the file arrives with text | Multifunction copiers, some standalone network scanners |
| On the attached computer | The scanner sends images to a desktop program, which adds text as it saves | Desktop sheet-fed scanners with bundled software |
| On a remote server | The device or its app uploads the pages, and a searchable file comes back or lands in cloud storage | Scan-to-cloud features, phone scanning apps |

To find out which applies to a machine in your office, run a simple test. Disconnect the scanner's computer from the network, or scan to a USB stick at the copier, and ask for a searchable PDF. If you get one, the recognition is in the device or the local software. If the option is greyed out or the job waits, a server is involved.

The answer matters for two reasons. The first is quality, since the recognition program, and not the glass, determines how well faint or small print is read. The post on [what an OCR engine is](https://www.schemalaw.com/blog/what-is-an-ocr-engine) explains that component and why two of them disagree. The second is confidentiality. If pages leave the building to be read, you need to know where they go and on what terms, exactly as you would for any online service.

### What the hardware does contribute

The scanner still sets the ceiling. Software can only read what the image contains, so a soft, skewed or over-compressed picture limits every program that touches it afterwards. A machine earns its keep by feeding paper straight, lighting it evenly and saving at a sensible resolution. The post on the [best DPI for OCR](https://www.schemalaw.com/blog/best-dpi-for-ocr) covers the resolution choice.

## What the searchable PDF button on an office copier does

The searchable PDF button tells the copier to run its own built-in recognition on each page after scanning, then save one file that holds the page pictures with the recognized text hidden behind them. Nobody reviews the text. The file reaches your inbox looking exactly like an ordinary scan.

Step by step, a typical device does this:

1. Scans each sheet to an image, using the resolution and color mode set on the panel.
2. Often applies its own adjustments: straightening, removing blank pages, compressing the image.
3. Passes each image to the recognition program stored in the device.
4. Writes the recognized characters into the PDF as invisible text.
5. Sends the file to the destination you chose.

Three things about that process are easy to miss.

**The settings are few.** A desktop OCR program may let you choose languages, handle columns or review doubtful words. A copier panel may offer little more than an on-off switch and a language list. If the default language is wrong for the document, the text suffers and nothing tells you.

**The option may not be there.** Whether a machine offers searchable PDF depends on its model and how it was configured. Two copiers on the same floor can behave differently. Scan the same page on each and compare before you assume.

**The result carries no warning.** A page the device read badly looks the same in the file as a page it read well. The only evidence is in the text itself. Open a scan from your copier, select a paragraph, and paste it into a blank document. Thirty seconds of reading tells you what your machine's recognition is like.

Paste one exhibit into a blank page and look for the party names, the date of the proceeding, and any obligation printed in small type. Those strings are the proof you will search for later. If the copier drops the obligation, you still see it on the picture and you miss it in the paste.

### A quick test for the copier you have

Take five pages that represent your real paper: a clean typed letter, a faxed page, a page with a table, a page with small-print footnotes and a page with a stamp or handwriting. Scan them once as a searchable PDF. Then, for each page, search for a surname, a dollar figure and one ordinary word, and paste the page's text into a document to read against the original. Keep the five results. They are your baseline for any other tool you try.

## Scanner OCR versus running OCR on the file afterwards

OCR at the scanner is convenient: one step, no extra software, and every scan arrives searchable. OCR afterwards is controllable: you keep a clean image, choose the tool, check the result, and can run it again later. The scan itself is the same in both cases. The difference lies in who decides how the text is made.

| Question | OCR at the scanner | OCR on the file afterwards |
| --- | --- | --- |
| Steps for the person scanning | One | Two |
| Choice of recognition tool | Whatever the device has | Any tool you select |
| Settings | Minimal | As many as the tool offers |
| Can you see the recognized text before it is saved? | No | Often yes |
| Can you redo it with a better tool? | Only if you kept or can recover a clean image | Yes, from the stored image |
| Output | Searchable PDF, sometimes little else | Searchable PDF, Word, text, spreadsheets, data |
| Works on files you received from others | No | Yes |
| What you must manage | Device settings | A second copy of each file |

The last two rows decide more than the rest. PDFs that arrive from other people's scanners never pass your copier, and no button on it helps with those. A firm needs a way to run OCR on files in any case. The real question is whether the copier should also do it for outgoing scans.

### A worked comparison

A paralegal at a fictional firm scans a 60-page set of bank statements for the matter of *Estate of Corinne Abernathy*.

*At the scanner.* She presses searchable PDF and the file is in the matter folder two minutes later. A search for "Abernathy" finds 40 pages. Nobody notices that the small-print transaction tables on the other 20 pages were read poorly, because the pages look fine.

*Afterwards.* She scans to plain image PDF, then runs the file through an OCR tool that shows the parsed pages on screen. On page 7 she sees a column of amounts has come out garbled, rescans that sheet at a higher resolution, and exports a searchable PDF plus a spreadsheet of the tables.

The second route took longer. It also produced a check, a correction and a second format, none of which the copier offered. Whether that is worth the time depends on what the documents are for.

## What is auto OCR and where does it run unasked?

Auto OCR is recognition that runs without anyone starting it. A setting, a scan profile or a storage system applies OCR to every file that passes through. It is useful, because nothing is left unsearchable by oversight. It also means text gets attached to documents that no person has ever looked at.

Places it commonly runs without a deliberate request:

- **Scan profiles on a copier.** Someone set "searchable PDF" as the default years ago, and every scan since has been through the device's recognition.
- **Bundled scanner software.** The desktop program that came with a sheet-fed scanner may add text to every file it saves unless told otherwise.
- **Phone scanning apps.** Many recognize text as soon as you photograph a page, so the PDF you share already has a text layer.
- **Cloud storage and email.** Some services read the images you upload so their own search can find them. The text may live in the service's index and not in your file.
- **Document management systems.** A system may run OCR on image files as they are filed, and either store the text alongside or replace the file with a searchable version.
- **Review and e-discovery platforms.** Processing may include OCR for files with no extractable text. The post on OCR before uploading to a review platform covers [whether hosted review platforms OCR your uploads for you](https://www.schemalaw.com/blog/ocr-before-uploading-to-a-review-platform#do-hosted-review-platforms-ocr-your-uploads-for-you).

### Three things to find out about any automatic step

1. **Does it change the file, or only build an index?** A system that keeps text in its own index leaves your PDF untouched. A system that rewrites the PDF gives you a different file from the one that came in.
2. **Does it run on files that already have text?** Running recognition over a PDF saved from a word processor can add a second, less exact copy of text the file already held.
3. **Where does the recognition happen?** On the device, on your server, or with an outside provider.

Auto OCR does not excuse the checking. A search across a folder of automatically recognized scans is a search of whatever each device and service happened to produce, with different tools and settings over the years. Treat the text as a finding aid until someone has compared a sample of it with the pages.

## Keeping an untouched image when the scanner adds text

Keep one copy of every scan as plain images with no text and as few automatic adjustments as you can manage, and treat the searchable version as a working copy made from it. If the scanner adds text and alters the picture in one pass, you may have nothing clean to go back to when a better tool comes along.

Adding hidden text does not, in principle, change the picture. The trouble is the company it keeps. A device asked for a searchable PDF often applies other processing in the same pass:

- straightening and rotating pages
- cropping to the detected paper edge
- dropping pages it judges blank, which can include the back of a sheet with a faint stamp
- converting to black and white
- heavy compression, to keep the file small enough to email

Each of these can be reasonable. Together they mean the file is the device's interpretation of the paper. For an original will, a signed agreement or a document whose appearance may be questioned later, you want the plainest capture you can get.

### A routine that keeps both

1. **Set up two scan profiles.** Name them so nobody has to think: "Archive image" (no OCR, moderate compression, no blank-page removal) and "Working searchable".
2. **Scan originals that matter with the archive profile.** Store that file as the master, and do not edit it.
3. **Make the searchable copy from the master.** Use the copier's own recognition or a separate tool. Give it a suffix such as `_ocr`.
4. **Check the count.** Confirm that the page count matches the paper, front and back.
5. **Note what was done.** A line in the file record, "scanned 4 Mar, archive profile; OCR copy made separately", answers questions years later.

If your matter records hash values, record the hash of the master. A copy with added text is a different file with a different value, even when every page looks the same.

If the only file you have is the copier's searchable PDF, you are not stuck. A separate OCR tool can read the page pictures again and produce fresh text. Be aware that the old text may stay in the file next to the new, so test a search afterwards. Schema Law's searchable PDF export, for example, is your original file with a hidden text layer added, while a Word or plain text export, available on every plan, holds the text as read from the pages. The [export formats section of the homepage](https://www.schemalaw.com/#formats) lists them.

## Which should a firm choose: OCR at the scanner or later?

Most firms should do both, on purpose. Let the scanner add text to routine paper where a rough finding aid is enough. Run OCR afterwards, with a tool you have tested, on anything you will quote, produce, file, calculate from or rely on. And always have a way to OCR files that other people scanned.

| Kind of scanning | Sensible default | Why |
| --- | --- | --- |
| Daily incoming post, filed for reference | OCR at the scanner | Speed matters more than perfect text |
| Closed-file backlog, kept for retention | OCR at the scanner or in bulk later, after a sample check | Volume is high and use is occasional |
| Exhibits and filings | Archive image, then OCR afterwards and check | The text and the image will both be relied on |
| Financial records, tables, forms | OCR afterwards, with table output | A copier's flat text layer loses the columns |
| Originals with evidential weight | Archive image as master; searchable copy separate | Appearance may be examined |
| PDFs received from others | OCR afterwards | Your scanner never touches them |
| Anything confidential on a scan-to-cloud device | Find out where recognition runs before using it | Pages may leave the building |

### Questions that settle it for your office

- What did the five-page copier test show? If names and numbers came through cleanly, device OCR may be good enough for routine work.
- Does anyone need Word, Excel or data from scans, or only search?
- Who will check recognized text, and when?
- Is there an untouched image for the documents that matter?
- For any tool that sends pages elsewhere, have you read its terms?

### If you are buying a scanner

Choose the hardware for how it handles paper: a reliable feeder, duplex scanning, even image quality, and profiles your staff can select with one touch. Treat bundled recognition as a bonus to be tested and not as the reason to buy. The panel settings that most affect recognition later are listed in the guide to [scanner settings that make OCR work](https://www.schemalaw.com/blog/how-to-digitize-law-firm-paper-files#scanner-settings-that-make-ocr-work).

Then pick the recognition tool separately, by running your own difficult pages through it. Schema Law's free plan takes one document a day of up to three pages, enough for a small sample, and the [pricing page](https://www.schemalaw.com/pricing) shows the page limits for larger files. Whichever tool you choose, the rule for the scanner stays the same: get a good picture first, and keep it.

## 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)