Adobe OCR: What Acrobat's Text Recognition Does

Adobe OCR is the text recognition built into Adobe's PDF products. Acrobat on a desktop adds a searchable text layer to a scanned PDF, the Adobe Scan phone app recognizes text as it saves a scan, and a free Adobe online tool does the same in a browser. Adobe does not list text recognition among free Acrobat Reader's features.

Each of those has its own controls and limits. People use "Adobe OCR" for all of them, which causes confusion in an office where one person has Acrobat Pro, another has only the free Reader, and a third scans with a phone. This post separates the products, using Adobe's help pages as the only source for what each one does. It makes no claim about how well any of them reads a page. Test that on your own documents.

What is Adobe OCR and which Adobe products include it?

Adobe OCR is optical character recognition inside Acrobat, the Adobe Scan mobile app and Adobe's online OCR tool. Adobe's help page on recognizing text in scanned documents says a scanned PDF contains only image data, and that Acrobat's OCR feature converts image text into selectable, searchable text.

The same page describes the result: Acrobat creates a searchable text layer in your PDF. The page still looks like the scan. What changes is that a search finds words and a cursor can select them.

Adobe describes four places where its recognition runs:

The second row surprises people. Adobe's page on editing scanned documents says that when you open a scanned PDF and select Edit, Acrobat automatically applies OCR and converts the file to an editable PDF copy. That is a different outcome from a searchable scan. A searchable scan keeps the page picture. An editable copy is a rebuilt page whose words you can retype, and for a signed or filed document that is rarely what you want.

For the online tool, Adobe's free OCR page says you select or drop a PDF, the tool applies OCR after the upload, and you sign in to download the searchable PDF or get a link to share it. The page says the free tool recognizes text in PDFs, and that recognizing text in image formats such as PNG, JPG and TIFF, or editing the recognized text, calls for Acrobat Pro.

So "adobe ocr pdf" can mean a desktop command, a phone app or a web page. Before you tell a colleague to "run it through Adobe", find out which of the three they have.

After the text layer is written, try a search for the plaintiff and the defendant in the caption, the damages figure, and the hearing date. Open the complaint page those words sit on and read the highlighted line. Edit a doubtful word only when the page image shows the same spelling.

Where the Recognize Text command sits in Acrobat

In current Acrobat the command is under All tools, in the tool named Scan & OCR. Adobe's instructions are to open the scanned PDF, select All tools, then Scan & OCR, then In this file, and finish with the Recognize Text button. A second entry, In multiple files, handles a batch.

Text recognition also appears in three other places in the desktop program, each documented by Adobe:

• While scanning. Adobe's page on scanning documents to PDF lists a Recognize Text (OCR) option among the scanner settings, so paper can become a searchable PDF in one pass.
• While cleaning up a scan. The page on enhancing scanned documents describes an Enhance scanned file command with a Recognize text checkbox.
• While editing. Adobe's page on editing scanned documents says that selecting Edit on a scanned PDF applies OCR automatically.

One more trigger is easy to miss. Adobe's page on fixing text issues in scanned PDFs notes that if OCR has not been applied and you try to select text or use Read Out Loud, Acrobat prompts you to run OCR.

Adobe's own pages do not all give the same path. Its article on how to make a PDF searchable describes a different one: head to Tools and select Recognize Text. If your screen does not match the steps on the current help page, look for the words Recognize Text. The click-by-click procedure for one file and for a folder is in the post on how to make a PDF searchable in Adobe Acrobat.

Which settings does Adobe let you change before recognition?

Adobe documents three choices before recognition: the pages to process, the language of the text, and the output type. The page range and language are in the first dialog. Adobe's scanning page puts the output type under a Settings button.

Pages

You pick a page range. This matters more than it sounds. A 300-page production that mixes typed pages and scans does not need recognition on every page, and limiting the range shortens the run.

Language

Adobe's scanned PDF settings page describes Document Language as the setting that sets the OCR language for character detection. If the file is a Spanish-language lease, set Spanish before you start. Recognition run in the wrong language is a common cause of garbled accents and broken words in any OCR program.

Output

The same settings page says Output selects the PDF type, and that all of the output options require an input of at least 72 dpi. Adobe's current page does not list the individual options by name or say what each does to the page image, so compare a page before and after on a copy.

Settings that apply while scanning

When Acrobat drives the scanner, more controls appear. Adobe's settings page lists these, among others:

Adobe's note on adaptive compression recommends scanning grayscale or color documents at 300 dpi and black-and-white documents at 600 dpi for that feature. Filters such as background removal change the picture of the page. For a document whose appearance matters as evidence, decide whether you want that before you switch them on.

Adobe Scan on a phone versus Acrobat on a desktop

Adobe Scan is a separate app with fewer controls. Adobe's settings page for Adobe Scan on iOS says the app converts image text in scans to actual text by default, using the language of your device. You can change the language or switch recognition off. The page describes no page range and no choice of output type.

The differences that affect legal work:

On the page limit, the iOS settings page says OCR does not run on scans longer than 25 pages for general users, and up to 100 pages with a paid subscription. The Android settings page gives 5 pages for general users and the same 100 with a subscription. A 40-page lease photographed in a client's office may therefore be saved as a PDF with no text in it at all. Check before you rely on it.

Adobe's troubleshooting page for the Scan app lists conditions that may impair recognition: a dark background, a source document in more than one language, a rotated document, and light that is too bright or too dark. It also says OCR produces no output when the selected language does not match the document. The same page lists a network connection and an Adobe Document Cloud connection as required.

That last point is the one to think about with client papers. Adobe's Android settings page says scans are saved to Adobe cloud storage as PDFs by default, so a phone scan is quick and also puts the pages in an online account. The post on how to OCR a photo of a document covers phone capture in general, including how to take a picture that software can read.

A short example. Paralegal Dana Whitlock photographs a 12-page handwritten and typed intake packet at a client's kitchen table using Adobe Scan on an Android phone, signed in with a free account. Twelve pages is over the 5-page figure Adobe gives for general users on Android, so she should not assume the saved PDF is searchable. Back at the office she opens the file in Acrobat, tries to select a word, and runs Recognize Text there if nothing selects.

Does Acrobat Reader recognize text in a scan?

Adobe does not list text recognition among the things Acrobat Reader does. Its Acrobat Reader page describes Reader as the free app for viewing, annotating and signing PDFs. The recognition steps in Adobe's help are written for Acrobat, and Adobe's plan comparison lists turning scans into searchable PDFs under a paid plan.

The plan wording is on Adobe's Acrobat plans page, and the post on making a PDF searchable in Adobe goes through which Adobe plan you need. If you are unsure what your licence includes, open a scanned PDF and look for Scan & OCR under All tools.

Reader still has a part to play once recognition has been done somewhere else. Adobe's Scan troubleshooting page says that working with converted text in a scan, meaning selecting it, commenting on it or marking it up, requires Acrobat Reader. In other words, Reader works with a text layer that already exists. It is the program for reading and searching a scan that someone else has made searchable.

That gives a person with only Reader three routes:

1. Use the online OCR tool in a browser, signed in with an Adobe account, for a PDF.
2. Use the Adobe Scan app for paper you can photograph, within the page limits.
3. Ask a colleague with a paid Acrobat plan to run recognition and send the file back.

A fourth route is a tool from another maker. The post on when to look for an Adobe Acrobat OCR alternative sets out when that makes sense. If what you need is the text or the tables in another format, such as Word, Excel or CSV, a converter is a different kind of tool from a PDF editor. Schema Law is one: you upload a PDF, PNG or JPEG in a browser and pick an export, and the export formats section of the homepage lists the eight choices.

What Adobe tells you to check after recognizing text

Adobe's advice is to review the document after running OCR to make sure the text is accurate and complete, and to correct errors by hand or run OCR again with adjusted settings. It also tells you to save a backup copy of the original scanned PDF before you start.

Both instructions are on Adobe's recognize-text page. A routine for keeping an untouched copy of the scan turns the second one into a habit.

Adobe gives you a tool for the review. Its page on fixing text issues explains that when a match is uncertain, Acrobat marks the word as a suspect: it shows the original image and places searchable text behind it. The Correct recognized text command in Scan & OCR, with the Review recognized text box ticked, steps through those suspects. For each one you see the scanned image and the suggested text, edit the Recognized as field if needed, and select Accept.

Two cautions about that review:

• It covers only the words Acrobat doubted. A word recognized wrongly with no doubt attached is not offered to you. A date read as 2019 when the page says 2018 can pass straight through.
• It corrects the text behind the picture. Adobe describes a suspect as the original image with searchable text placed behind it. The field you edit is that text, which is what search and copy use.

For legal documents, add your own checks to Adobe's. Start with the two-minute test that Acrobat added searchable text. Then read every party name, date, dollar amount and section number against the page image, and confirm the page count matches the original.

If Acrobat will not work on the file at all, or the page image is too poor to read, see what to do when Acrobat refuses to recognize text.

The post on why OCR output is full of errors explains which mistakes come from the scan and which from the software, and how to proofread efficiently. None of this is special to Adobe. Every recognition tool can misread a page, and the person who signs the filing is the one who has to catch it.