Do Your Existing PDFs Need OCR or Rescanning?
If an existing PDF contains clear page images but lacks usable searchable text, assess OCR before arranging another scan. If essential information is missing from the image, first look for a better digital copy or assess the available paper originals. These are different problems and may require different work.
A folder labelled “scanned records” can contain several kinds of PDF. Some already have useful text, some contain page images only, and others have recognised text that does not match what a reader sees. One file can also contain different page types. Start by describing the problem in representative files rather than requesting the same treatment for the entire archive.
Check the image and the text separately
Open a working copy in an approved PDF application. Keep the received files unchanged so that any proposed output can be compared with them. Adobe's text-recognition guidance explains that Acrobat's OCR can add a searchable text layer to an existing scanned PDF and recommends keeping a backup before editing.
For selected pages, ask two questions:
- Can a person read the required information in the page image? Look at the particular names, reference numbers or passages staff need. Record missing edges, unreadable areas and absent pages.
- Can the application find and copy that information correctly? Search for a known visible term, then compare a short copied passage with the page. A successful search is useful evidence for that term, not proof that every word is correct.
Include different sections, layouts and known problem pages. A selectable cover sheet does not tell you whether a scanned appendix has useful text.
Choose the next step for each problem
This table is a buyer's assessment framework, not a diagnosis or a promise that every file can be repaired.
On a small screen, scroll the table sideways if needed.
| What you observe | Work to discuss | Evidence to request before a wider batch |
|---|---|---|
| Required text already searches and copies correctly in the pages checked | Whether those pages need any text processing | Confirmation of what will be left intact and what still needs assessment |
| Page image is readable, but required text cannot be found, selected or copied | Check whether usable text is absent and whether restrictions affect the test; assess OCR if appropriate | A processed sample with agreed searches and copied-text comparisons |
| Page image is readable, but copied words or references are wrong | Assess the text layer and whether correction or replacement is appropriate | Before-and-after checks of the specific errors, plus any remaining exceptions |
| Required information is cropped out, missing or unreadable in the image | Look for a better digital version; assess rescanning originals if available | A source comparison showing whether the missing detail can actually be captured |
| Search works inside the PDF but not in the archive application | Investigate the destination's search and indexing separately | The same known-term search performed in the intended application |
For the separate decision about searchable files versus spreadsheet fields, use the searchable PDF and data extraction guide. The decision here is what condition the existing files are in before choosing the work.
Set aside signed or restricted files
Keep digitally signed or restricted files out of a routine processing batch until their treatment is agreed. For example, OCRmyPDF's signed-PDF documentation states that it cannot both preserve a PDF's digital signatures and add OCR. Do not assume every file can be altered in the same way.
If selection or copying is restricted, record that constraint for the document owner to resolve; do not interpret it as proof that OCR is missing. Use only an approved processing environment for confidential records, rather than uploading them to an unapproved online converter to test a file.
A readable image can still have incorrect recognised text
Suppose a scanned page with an OCR text layer visibly shows the fictional reference BX-1048, but copying it produces BX-1O48, with a letter O replacing a zero. That points to a text-recognition issue in the checked location. It does not, by itself, establish a need to rescan the paper.
An exact search for BX-1048 could miss this reference even though the page has a text layer. A failed search alone does not show that OCR is missing.
Adobe documents an Acrobat process for reviewing words its OCR flagged as uncertain against the scanned image. An error the software was confident about may not be flagged, so check the references that matter directly.
OCRmyPDF's existing-OCR guidance distinguishes replacing an OCR layer from forcing every page back through image conversion, and notes that some existing OCR cannot be detected or replaced. Its forced conversion can reduce quality or lose vector content; it is not equivalent to replacing only the recognised text.
Ask the provider to explain which pages it proposes to process and how it will treat usable text already present. Avoid treating “run OCR again” as a complete specification. Compare a sample before deciding whether correction, replacement recognition or another approach is appropriate. These software examples explain possible approaches; they do not identify Pexatech's processing tools or guarantee a result.
When the source image needs attention
A second recognition pass cannot reveal an entire line that was cut off and never captured in the supplied file. Search for an earlier, clearer digital version before arranging access to physical originals.
Other problems need assessment rather than an automatic rescan. Tesseract's image-quality guidance describes how noise and skew—such as speckled backgrounds and crooked pages—can affect recognition. This supports testing the input and proposed processing; it does not establish a universal setting that will fix every archive.
If rescanning is proposed, ask which original pages are needed and how they relate to the existing files. Record unavailable originals and unresolved gaps. For bound material, the bound-document guide covers handling questions that must be settled separately.
A fictional mixed-folder assessment
This example is invented to illustrate a decision. It is not a customer project, tested OCR result or service commitment.
An office coordinator reviews a small set of files from an old project folder. One PDF has usable exported text, another has readable scanned pages with no useful search result, and a third has an appendix whose right-hand edge is missing. A fourth finds the expected reference when opened directly but produces no result in the team's archive search.
The coordinator records four different next steps: assess whether the exported file needs processing at all; request an OCR sample for the readable scan; locate a better source for the cropped appendix; and ask the archive administrator to investigate the search behaviour. The brief identifies these groups instead of asking for the whole folder to be rescanned.
Send a focused assessment request
An initial enquiry can describe the collection without attaching confidential records. Include:
- The approximate number of files and pages, how they were created if known, and the languages present.
- What people need to find, with a general description of the failing search or copying behaviour.
- Whether the issue affects selected pages, whole documents or an unknown portion of the collection.
- Whether better digital copies or paper originals are available.
- The destination application, required output and any restrictions on processing the files.
- A request to agree representative samples, review responsibilities and treatment of unresolved exceptions before a wider batch.
Pexatech's OCR service overview provides a starting point for that discussion. Confirm whether the proposed work fits the files and requirements before sharing them. Once the approach is agreed, the scanning quality checklist can help define delivery checks; PDF versus PDF/A addresses the separate output-format decision.
Frequently Asked Questions
Should we print existing PDFs and scan them again?
Start with the digital files. If their page images are usable, OCR may be assessed directly on those files. If important detail is missing, look for a better digital source or the available originals. Printing the same incomplete PDF does not recover information that was absent from it.
Does selectable text mean a whole PDF has usable OCR?
No. Selection on one page does not establish the condition of other pages, or the accuracy of the selected text. Check representative sections and known problem pages. Compare copied text with the visible page, and record where the result is missing, incorrect or incomplete.
What if the paper originals are no longer available?
Ask whether a better digital version exists and have the available files assessed. Some recognition problems may be improved, but missing page content needs an appropriate source. Record unresolved gaps and agree how they will be identified in the output; do not present an inferred replacement as recovered original text.
Discuss the PDFs your team already has
Describe the files, the search problem and the output you need. Confirm suitability, scope and sharing arrangements before sending documents or confidential samples.
Discuss Your OCR Requirements