A searchable PDF is more than an image with a text layer. Use this reusable quality checklist to test OCR accuracy, confirm that every page and file is present, and decide whether a scanned document batch is ready for storage, search, accessibility review, or system migration.
Overview
OCR, or optical character recognition, converts words in a scanned image into machine-readable text. When it works well, users can search for names, dates, clauses, account numbers, and other content without reading every page. A searchable PDF normally keeps the original page image while adding an invisible or selectable text layer.
That distinction matters during a document digitization project. A PDF can open successfully and still contain missing pages, incorrect text, poor page order, incomplete bookmarks, or an OCR layer that is too inaccurate for the intended use. Visual quality and text quality should therefore be checked separately.
Before accepting a batch from a document scanning service, define what “complete” means for your records. Confirm the expected page count, file structure, naming convention, resolution, color requirements, OCR language, metadata, and exception process. Requirements for a reference archive may differ from those for a searchable legal file, a medical record workflow, or a document management system migration.
For preparation guidance, see How to Prepare Documents for Bulk Scanning. If the project includes sensitive records, pair this review with a document scanning vendor security checklist covering custody and data protection.
Checklist by scenario
For every scanning project
- Confirm the inventory: Compare delivered files or page counts with the approved manifest, box list, folder list, or batch report.
- Check page order: Sample the beginning, middle, and end of each batch. Look for reversed pages, duplicated pages, blank pages, and inserts placed in the wrong file.
- Open representative files: Test PDFs from different document types, sizes, conditions, and operators rather than checking only the first file.
- Review image quality: Look for legible text, adequate contrast, correct orientation, clean edges, and acceptable handling of color, photographs, stamps, and handwritten notes.
- Test search: Search for a word visible on the page, a number, a date, and a name. Confirm that the result highlights the correct location.
- Check copying: Copy a short passage into a text editor. Look for missing spaces, merged words, unexpected characters, and errors in punctuation.
- Verify file integrity: Make sure files open without warnings and that page thumbnails, navigation, and downloads work as expected.
For bulk document scanning services
- Require a batch identifier that connects digital files to the source box, folder, or manifest.
- Check file naming against the agreed pattern. Names should be predictable enough for sorting, import, and later retrieval.
- Confirm that multi-page documents remain together and that separate documents were not combined.
- Review whether blank-page removal was allowed. A blank page may be intentional, especially when it separates sections or preserves the original record.
- Test folder structure, metadata fields, and index files if the output is going into a document management system.
- Ask how unreadable pages, unusual formats, torn pages, folded documents, and handwritten material were identified and reported.
For legal, medical, and compliance-sensitive records
- Preserve the original page image even when OCR text is available. OCR should support discovery; it should not silently replace the source image.
- Check identifiers carefully. Names, dates, case numbers, patient numbers, dollar amounts, and dosage information deserve targeted sampling.
- Confirm that exceptions are listed rather than quietly omitted. The report should identify pages that could not be scanned, recognized, or matched to a record.
- Review access controls, transfer methods, retention instructions, and chain-of-custody documentation separately from OCR quality.
- Where records will support a regulated workflow, have the responsible records, legal, or compliance owner approve the acceptance criteria.
Healthcare teams can use this checklist alongside medical record scanning requirements. The operational standard should reflect the use of the record, not simply the capabilities advertised by a vendor.
For migration into a business system
- Test the exact import format, including PDF version, file size limits, character encoding, metadata fields, and folder rules.
- Verify that bookmarks, hyperlinks, page labels, and attachments behave correctly after import.
- Run searches inside the destination system, not only in a desktop PDF application.
- Test duplicate handling and confirm what happens when a file fails validation.
- Keep a small, approved test set available so future software or workflow changes can be compared with the original result.
What to double-check
Use the following scorecard for each sample batch. Mark each item as Pass, Fail, or Review; add a note for every failure.
| Area | Acceptance question | Evidence to record |
|---|---|---|
| Completeness | Are all expected documents and pages present? | Manifest comparison and exception report |
| Image quality | Can a person read the page at normal viewing size? | Sample filenames and page numbers |
| Resolution and appearance | Does the output meet the project’s agreed image specification? | Representative files and specification |
| Deskewing and orientation | Are pages straight, upright, and consistently framed? | Visual sample review |
| OCR accuracy | Do searches and copied text match the visible source? | Test terms, error notes, and sample pages |
| File structure | Are document boundaries, page order, and folders correct? | Batch map or import log |
| Bookmarks and navigation | Can users move through long files efficiently? | Bookmark and page-label test |
| Names and metadata | Are filenames and index fields complete and consistent? | Metadata export or validation report |
| Accessibility | Can the intended users navigate and read the file with available tools? | Text selection, reading order, and tag review |
| Exceptions | Are damaged, unreadable, or unrecognized pages documented? | Exception list with disposition |
Do not rely on a single overall OCR percentage. Accuracy can vary substantially by typeface, print quality, handwriting, skew, language, tables, stamps, and background noise. Instead, create a test set that reflects the real batch. Include clean typed pages, faint pages, forms, tables, copies, handwritten annotations, and the document types that matter most to your operation.
For each category, define an action. For example, a minor OCR error in a historical reference file may be recorded as an exception, while an incorrect account number in a transaction record may require reprocessing or manual indexing. The acceptance decision should be based on business risk and intended use.
Also verify that “searchable” means searchable in the way your users need. Test exact names, partial terms, numbers, punctuation, accented characters, and common abbreviations. Search failures can result from OCR errors, a missing text layer, an unexpected language setting, or limitations in the destination application.
Common mistakes
- Checking only whether the PDF opens. Opening a file proves little about completeness, OCR accuracy, or correct indexing.
- Testing one perfect sample. A clean first file may not represent folded pages, carbon copies, forms, photographs, or later batches.
- Confusing selectable text with accurate text. Text can be selectable and still contain critical recognition errors.
- Using visual review alone. A page may look clear while its OCR layer contains missing words or incorrect numbers.
- Accepting unnamed exceptions. If a page cannot be recognized or scanned, the issue should be visible in a report with a defined next step.
- Skipping destination-system testing. A file that works on a desktop may fail during import because of naming, metadata, size, or compatibility rules.
- Overlooking accessibility. Searchability does not automatically guarantee logical reading order, useful navigation, or compatibility with assistive technology.
- Changing standards mid-batch. Agree on requirements before production and document any approved changes so results remain comparable.
When a batch fails, avoid requesting a vague “quality improvement.” Identify the affected files, page numbers, error type, sample evidence, and required remedy. Depending on the issue, the remedy may be rescanning, OCR reprocessing, manual indexing, file splitting, metadata correction, or acceptance with a documented exception.
When to revisit
Review this checklist before every major digitization batch, vendor acceptance review, system migration, or recurring quality audit. Revisit it when document types change, a new OCR or PDF tool is introduced, the destination repository changes, or users report that searches are unreliable.
It is also useful to review the checklist before seasonal planning and budget cycles. Confirm that your sample set still reflects current records, that file naming and metadata rules remain compatible with your systems, and that the team still agrees on which errors require rework.
Keep a dated copy of the approved requirements, test set, scorecard, exception log, and final acceptance decision. That record makes later comparisons easier and helps distinguish a scanning problem from a change in software or workflow.
- Write the acceptance criteria before production starts.
- Approve a representative test batch.
- Run both visual and text-based checks.
- Record failures by file and page, not only as a general comment.
- Test the output in its intended storage or business system.
- Accept, remediate, or document exceptions before releasing the next batch.
For project planning, compare this quality review with a document scanning cost and pricing template so rework, indexing, and exception handling are considered alongside the initial scan. If you are still selecting a provider, use the document scanning service vendor-vetting checklist to ask how quality is measured before files are delivered.