PDF Redaction Use Cases
Redaction is required wherever a document leaves an organization with some of its content removed: public records and FOIA responses, healthcare records with PHI, financial statements and KYC files, legal discovery and privilege review, and documents bound for an AI pipeline. Each has its own detection rules and its own evidence requirement.
Public Records and FOIA Responses
A records request arrives with a deadline and a list of exemption categories. Every released page has to lose the exempt content, and a black box drawn over text does not remove it. Redaction destroys the underlying text, image and vector content instead of hiding it.
In the Apryse Web SDK, your reviewers mark text or regions in the WebViewer UI, and the browser applies the removal. In the Apryse Server SDK, your code finds matches through text search or text extraction and applies them on your servers, with no viewer involved. Read the full workflow in FOIA and public records redaction. If your case management runs on Appian, see case management redaction in Appian:
Healthcare Records and PHI
Clinical records and claim files carry PHI identifiers on every page, and each release has to remove them. Your team writes the string, wildcard or RegEx rules that match those identifiers, and redaction removes the matches permanently.
In the Web SDK, a reviewer marks text or regions in WebViewer and applies the removal in the browser. In the Server SDK, rules of the same kind run across claim files on your servers, so PHI stays in your environment. Redaction supports your HIPAA workflow, and your app owns the compliance program and the accounting of disclosures. The browser path is covered in HIPAA redaction in the browser:
Financial Services: Statements, KYC and Account Data
Statements and KYC files carry account numbers, IBANs, card numbers and customer identifiers, and they arrive in volume. Your team writes the pattern for each identifier, and redaction matches it with RegEx, wildcards or plain strings.
The Server SDK applies the removal on your servers, and your code owns the loop over the files. The Web SDK serves the analyst who marks a document by hand in WebViewer. Redaction removes image and vector content inside a marked region as well as text. Both surfaces remove the content permanently, so the removed data cannot be recovered from the file. Your app keeps the retention record. The detail is in financial services document redaction.
Legal Discovery and Privilege Review
A discovery set arrives in mixed formats, and a missed privileged passage creates real exposure. Privilege calls stay with your reviewers. In the Web SDK, they mark passages and regions in WebViewer. In the Server SDK, your code searches the set for terms or patterns and marks matches for removal on your servers, with no upload to a vendor cloud.
Marked regions stay movable and redefinable until removal runs, so a reviewer corrects a call before it becomes permanent. The removed content is destroyed, not hidden. Your app keeps the privilege log. Scanned pages need OCR first. For the operations view, read redaction for legal operations.
Removing PII Before AI and RAG Processing
Documents bound for a model leave your control when a pipeline sends them to a service you do not run. Redaction removes personal data before that hand-off. Your pipeline marks matches through text search or text extraction, the Server SDK applies the removal on your servers, and the cleaned file moves on to retrieval-augmented generation (RAG) or any other model step.
Matching uses string, wildcard and RegEx rules, such as SSN and credit card patterns. The Server SDK runs inside your infrastructure, so the unredacted file never reaches a third-party service. Detecting PII is not removing it, and a flagged file still carries the data. Pipeline details are in PII detection before AI and RAG.
Your app records the evidence listed above. The SDK removes the content. That removal step stays constant: redaction permanently removes visible text, images and vector content, and it cannot be undone once applied.
Hidden content such as metadata and embedded scripts falls under Sanitization, a separate capability in the base Server SDK package. Neither replaces the other.
Scanned pages have no text to search, so they need OCR first. See redacting scanned documents. For the full capability, see the PDF redaction library. For the sequence a compliance review follows, see the compliance redaction process.
Where Each Workflow Runs
Choose the surface by who is in front of the document. When a person marks and reviews, the Web SDK runs redaction in the browser through WebViewer. When no one does, the Server SDK runs it on your servers. Redaction needs an add-on license on both.
The Server SDK includes no work queue, scheduler or batch orchestration, so your code owns the loop. See redacting documents in bulk. For regulated-industry requirements, read the enterprise-grade redaction capability. Apryse is an SDK, not an end-user redaction application, so your team builds the workflow around it.
What Changes Between Use Cases
Three things change between use cases: what has to be found, which surface runs the workflow, and what evidence has to survive. The table maps each use case to all three and to its deeper guide:
Frequently Asked Questions
The most common are public records and FOIA responses, healthcare records containing PHI, financial statements and KYC documents, legal discovery and privilege review, and PII removal before documents reach an AI pipeline. Each needs a different detection rule set. The table above maps each one to its surface and its deeper guide.
What has to be found. Healthcare workflows match PHI identifiers across clinical records and claim files. Financial workflows match account numbers, IBANs, card numbers and customer identifiers across statements and KYC documents. The removal step is the same, and your app owns the retention rules.
Not by pattern, because a scan has no text to match. Run the Server SDK OCR Module first, and the recognized text is what your rules match against. Marking a region by hand still removes the image content inside it. OCR is a separate add-on license, and Redaction needs the Redaction Package on the license key.
Workflows where a person marks and reviews redactions run in the browser, through WebViewer on the Web SDK. Workflows with no person in front of the document run on your servers with the Server SDK. Redaction is an add-on license on both. The Server SDK includes no queue or scheduler, so your code owns the loop across files.
Yes, if the documents contain personal data and the pipeline sends them anywhere you do not control. Detecting PII is not the same as removing it, and a flag alone leaves the content in the file. Redaction removes it permanently before the hand-off.
