Adobe’s Scan & OCR example shows a suspect-word correction. Review flagged words and separately verify critical values; this example is not a result from your document.
A scanned PDF can show perfectly readable words while containing only a picture of the page. To make that text searchable, use Acrobat desktop’s Scan & OCR tools on a working copy, choose the correct document language, and then test the saved result. OCR recognition creates a text layer; it does not certify that names, amounts or reading order are correct.[1][2]
This guide covers the desktop Acrobat workflow when the signed-in account has access to Scan & OCR. Free Acrobat Reader can search text already present in a PDF; opening an image-only scan in Reader does not itself perform this workflow. Browser viewers and mobile scanning apps have different controls, and some capture apps may already have added recognized text.[4]
Check whether OCR is actually the missing step
Keep the received file unchanged. In a separate working copy, search for a distinctive word that is visibly on the page, then try selecting a short line and pasting it into a plain-text editor. Repeat on a second page: a PDF can mix scanned pages with pages that already contain text.
| Observed result | Likely next step | Acceptance check |
|---|---|---|
| Nothing can be selected; a visible word is not found | Treat the page as an image-only OCR candidate. | Recognized text can be searched after saving and reopening. |
| Text selects, but pasted characters or numbers are wrong | Review the existing recognition before running a whole-file batch. | Critical words and values match the visible original. |
| Some pages work and others do not | Identify the affected page range. | Each page type is tested, including inserts and rotated pages. |
| Copying or editing is blocked | Check authorization and document restrictions with its owner. | Use an authorized working copy; do not bypass restrictions. |
PDF permissions can restrict copying or changes independently of whether a text layer exists. A signed or locked document also needs a separate decision: request an authorized unsigned source rather than casually altering the signed record.[5][6]
Run a small recognition pass before processing the batch
- Save a clearly named working copy in an approved location and choose a representative page range.
- Open All tools > Scan & OCR, then choose In this file.
- Set the page range and recognition language. Open Settings if the job needs a different output option.
- Select Recognize Text, then save the result under a new name.
Adobe also provides a multiple-file workflow with output-folder and naming choices. Prove the sample first so the same language or naming mistake is not multiplied across a folder.[1]
Keep a short processing note: original filename, processed range, selected language, output filename and reviewer. For a mixed-language document, test representative passages from each language rather than assuming one successful English heading proves the entire file.
Fix the scan before blaming the recognition engine
A tilted, faint or patterned scan gives recognition less reliable input. Acrobat’s scan settings include deskew, descreen, background removal and sharpening; availability and effect depend on the scan and chosen settings. Compare a sample after enhancement instead of repeatedly compressing the only copy.[3]
| Page problem | Useful sample test | What to compare |
|---|---|---|
| Lines lean across the page | Try deskew on a copy. | Straight text and preserved page edges. |
| Printed dot patterns interfere with characters | Test descreen on a representative scan. | Legibility of small letters and punctuation. |
| Text is faint against an uneven background | Compare a modest background/sharpening adjustment. | Decimal points, thin strokes and shaded fields remain visible. |
| Source is clipped, blurred or missing content | Obtain a better scan or source document. | All required content exists before recognition. |
Do not accept a smaller file merely because its pages still look plausible at full-page zoom. Inspect small type and critical figures at a readable scale. A missing decimal point in the image cannot be repaired by a confident-looking text layer.
Review recognized text, then check what was not flagged
In Scan & OCR > Correct recognized text, enable Review recognized text. Acrobat presents suspect words with the image and recognized value; correct the value and accept it before continuing. This review is a starting point, not an exhaustive error list.[2]
Independently inspect the fields that drive the business decision: account identifiers, dates, totals, names, page references and table headings. Paste a representative paragraph and a table row into plain text. Look for a zero read as O, a one read as I, missing minus signs, joined columns or text read in the wrong sequence.
Worked example: a hypothetical invoice is searchable by the supplier name, but its invoice number copies with O instead of 0. The search test passed; the identifier check failed. Correct the recognized value where supported, compare it against the image, save, reopen and repeat both tests. Do not let the OCR output silently become the accounting source of truth.
Accept the saved file for a defined use
- Close and reopen the actual output file; search a known phrase on each representative page type.
- Copy critical sample values and compare them with the original image.
- Confirm page count, orientation and visible content are intact.
- Record unverified pages or fields and the person responsible for resolving them.
- Retain the original according to the organization’s document policy.
A searchable PDF is not automatically an accessible PDF. Tags, reading order, alternative text and an assistive-technology review are separate concerns. Likewise, searchable text is not a reason to send a sensitive document to an unapproved online conversion service.
If the next step is disclosure, follow the Acrobat redaction and hidden-information checks; OCR can make words easier to find but does not remove them. If the saved file looks correct on screen but fails on paper, use the PDF printing isolation guide instead of rerunning OCR without evidence.
Fit OCR into the document workflow
For a recurring scanning problem, start at capture: document language, legibility, page completeness, filenames and the destination folder. A consistent source scan is easier to validate than a pile of independently repaired files.
For capture-side device checks, see Brother printing and scanning troubleshooting. For legal-document intake, law firm document preparation for an AI pilot explains why document quality and human review remain separate acceptance steps.
Adobe licensing, installation and workflow configuration can be discussed through Adobe Support or Software Support. Bring the sample-test results and a suitably protected example, rather than assuming every failed search needs a new scanner.
Frequently asked questions
Why can I read a scanned PDF but not search its words?
The page may be an image without a recognized text layer. Test a known visible word and try copying a short line. Use OCR on an authorized working copy if the page contains only an image.
[1]Does opening the scan in free Acrobat Reader run OCR?
Should I run OCR again if my scanning app already recognized the text?
Which language should I choose for Acrobat OCR?
Why is text selectable but still incorrect?
Selectable text can contain recognition mistakes or an unhelpful reading sequence. Compare copied names, identifiers, punctuation and values with the original page before using them in another system.
[2]Does Correct recognized text find every OCR error?
No. Review the suspect words Acrobat presents, then independently verify the fields and passages that matter to the task. An unflagged word or number can still be wrong.
[2]Can I recognize only the scanned pages in a mixed PDF?
Yes. The In this file workflow lets you choose a page range. Identify the affected pages first, then test the saved output for each page type.
[1]What should I do if the scan is signed or copying is restricted?
Will OCR remove confidential information from a PDF?
No. OCR recognizes text; it does not redact it. Use the separate redaction workflow and verify the exact disclosure copy before sharing.
[1]Does searchable text prove that the PDF is accessible?
No. Searchability tests one aspect of the document. Reading order, document structure and other accessibility requirements need their own review.
[4]
