How to Prepare PDFs for a Public Records Response

A public-records response is not just a redaction job. It is a controlled transition from collected source material to a release copy that another person can understand and audit. The legal rules, exemptions, deadlines, and review requirements depend on the authority and jurisdiction. This guide covers the PDF handling sequence, not the decision about what must be disclosed.

Start with the request and a protected master

Before editing, write down the request scope in plain language: date range, people or departments, record types, and any agreed clarification. Give every collected source a stable identifier. Preserve those source files unchanged in the records system used by your organisation.

Make a separate working set for review. This distinction prevents a common failure: someone removes pages or marks text in the only copy, then cannot show what was collected or why the released file differs.

Your working log can be simple:

Field	Purpose
Source ID	Connects a release page to the preserved record
Date and record type	Helps reviewers find duplicates and out-of-scope items
Responsive pages	Records what addresses the request
Redaction decision	Names the approved reason without copying sensitive text
Reviewer and date	Shows who completed the required check
Release filename and page range	Reconciles the log with the file sent

Do not place confidential values in an informal spreadsheet merely to describe them. A reference such as "account number in paragraph two" is usually safer than repeating the number.

Separate responsiveness from redaction

First decide whether a record or page responds to the request. Then review the responsive material for authorised withholding. Combining those decisions makes it easy to exclude a whole page because one line is sensitive, or to redact a value simply because a detector found it.

Automatic detection can help locate repeated structured data. The PII review tool proposes identifiers such as email addresses, phone numbers, and account-like values across a document. Every hit still needs a person to decide:

• Is this value actually the type the detector suggests?
• Is it responsive to the request?
• Is there a current, approved basis for withholding it?
• Would a narrower redaction preserve more useful context?
• Does the same value appear in a header, attachment, image, or later page?

Detection reduces missed occurrences; it does not supply legal authority.

Handle scanned pages before detailed review

Scanned correspondence and old forms may contain only page images. Search will not find a name that has no text representation. Make the scan searchable before using text search as a completeness check, and proofread recognition results against the page where exact names and numbers matter.

If a scan is skewed or photographed at an angle, correct the image first. Do not treat an OCR result as the official record or silently replace the preserved scan. It is a review aid that makes long files easier to search.

For handwritten notes, maps, diagrams, or poor copies, assume manual page review will still be required. A search-based workflow is only as complete as the text it can read.

Apply real redaction to a release copy

Use the redaction tool for approved content. A black rectangle placed by a general editor is not enough because the original text can remain selectable underneath. The released output must remove the covered content and then be checked independently.

Work in deliberate passes:

1. Redact exact repeated values approved for withholding.
2. Review contextual items that a pattern cannot understand, such as narrative descriptions or signatures.
3. Inspect headers, footers, marginal notes, and repeated attachments.
4. Reopen the output and search for the full value and meaningful partial forms.
5. Confirm that the page still communicates the non-withheld context.

Keep an unredacted master under the appropriate access controls. Do not rely on undo history inside a working PDF as the preservation mechanism.

Check what the page view does not show

Public release creates a broader audience, so hidden material deserves its own pass. Run the privacy risk scanner against the redacted output. Review document properties, comments, form values, embedded attachments, links, hidden text, and other reported structures.

This is where a narrow redaction can fail operationally. A name removed from page five may still appear in the document title, a sticky note, or an attached spreadsheet. Flattening comments can make them visible and permanent; it does not decide whether those comments should be released. Remove or retain each item according to the approved disclosure decision.

The definition of PDF sanitization explains why metadata cleanup, flattening, encryption, and redaction are separate steps.

Assemble a release copy that can be referenced

When the review is complete, assemble records in the order stated by the response log. Use merge PDF only if one combined file is appropriate; some processes require separate files by source or exemption.

Add stable page numbers to the release copy when the authority's process permits it. A continuous range makes it possible for a requester, reviewer, or appeals officer to identify a page without guessing which source filename it came from. Do not cover existing content or change an official identifier already printed on the record.

Use filenames that describe the response and part number without exposing a withheld person's name. If the portal has size limits, split at logical record boundaries rather than producing arbitrary chunks that destroy context.

Validate the exact file being sent

A second-person review may be required by policy. Even when it is not, use a final checklist:

• Page count and order match the response log.
• Every redaction is visible in the expected place and leaves no selectable underlying value.
• Extracted text contains no approved withheld term or partial identifier.
• The hidden-data scan has been resolved or documented.
• Attachments and comments are present only when intentionally released.
• Page numbers and filenames reveal no excluded information.
• The file opens in a second reader and the intended pages print correctly.

Record the final filename, byte size, review date, and a fingerprint if your process uses one. Then send that exact copy. The strongest PDF workflow cannot repair an incorrect disclosure decision, but it can make the approved decision much harder to lose between review and release.