How to redact a PDF and check the saved result

In short

• A black rectangle can leave the words underneath recoverable. Use a redaction tool, apply the changes, then check the saved PDF.
• The free Starlight Tools route below worked on our three-page fictional test. Search found text, but we had to cover text inside pictures manually.
• Its downloaded PDF turned every page into an image. That removed searchable text and interactive features, and made our small file about six times larger.
• Keep the original separately. Check the downloaded copy before sharing it, including what remains visible around each black mark.

We tested with invented codes, pictures, a form and hidden document information. This is a practical check of one output, not a security audit or a guarantee for every PDF. For confidential work, use a tool your organisation approves and follow its release process.

1. Choose a tool that fits the document

If you use a Mac and only need to remove selectable text, check Redaction Selection in Preview's Markup toolbar first. Apple's instructions say the redaction becomes permanent when you close the document. We did not run the test file through Preview. Apple's Preview guide (opens in a new tab)

LibreOffice documents another free route through Tools > Redact, followed by Redacted Export (Black) or Redacted Export (White). Its ordinary Direct Export to PDF makes an in-progress review copy. That is a different command from final redacted export. This alternative is based on its documentation, not our trial. LibreOffice's redaction instructions (opens in a new tab)

For the browser walkthrough, open Starlight Tools' PDF redaction tool (opens in a new tab). We completed the trial without payment or an account. Starlight says processing stays in your browser. We did not audit that claim or its network traffic. Its privacy policy (opens in a new tab) also describes technical logs, advertising and analytics.

2. Open a copy and find repeated text

Keep an untouched original in a separate folder. Choose a working copy with Choose PDF.

1. Under Find and redact, enter a name, phrase or identifier in Exact word or phrase.
2. Select Find matches.
3. Review the page numbers and snippets. Select the intended matches, then choose Mark selected. The button may include the number selected.
4. Repeat for other details, including spelling variations and identifiers repeated in headers or footers.

In our test, a header code appeared on all three pages. Search found three matches, and we marked each one. We also found a body-text code and a hidden text layer behind a picture.

A page can contain visible words that the tool cannot read as text.

Enlarge image

Starlight search for RASTER-ORCHID-263 returning zero matches.

×

Search returned no matches for the fictional picture-only code. Inspect image content manually too.

Open full-size image

Enlarge image

The fictional code RASTER-ORCHID-263 visible inside a PDF image.

×

The same fictional code is visible in the page image, even though search found no matches.

Open full-size image

3. Cover pictures and check every page

Use Next page and Previous page to inspect the whole PDF. Drag across each image area that needs removing. Leave enough space around the content to cover the edges of every character.

Our picture-only code, RASTER-ORCHID-263, returned zero search matches while it remained visible on the page. We covered that image manually. A second picture had matching hidden text that search could find. The resulting redaction rectangle fitted closely around the text, so we added a larger manual mark over the complete image region.

Check filled form values and any visible comment symbols too. Review the surrounding content: a job title, address fragment or nearby sentence may still reveal the detail you meant to remove.

Use Zoom to inspect coverage. Our trial used black marks. Do not judge the result from a small thumbnail or assume that a search-selected rectangle covers everything you can see.

Enlarge image

A black manual redaction selected with a blue outline and a lower-right resize handle.

×

The selected mark has a resize handle at the lower-right corner.

Open full-size image

4. Apply the marks and download the result

1. Set Preview and output color to black if it is not already selected.
2. Under Output resolution, choose Balanced at 200 DPI to match our test.
3. Select Review and export. Check the number of pages and marks against what you intended to cover.
4. Select Apply redactions and create PDF.
5. Select Download redacted PDF. Locate the saved file, which had "-redacted" added to its name in our trial.

Enlarge image

Starlight output settings showing full opacity and Balanced 200 DPI, with rasterization tradeoffs.

×

Open full-size image

Our output contained 12 marks across three pages. It grew from 37,875 bytes to 231,306 bytes. Other files can differ substantially.

The result screen includes a Sanitization inspection. Read it, but continue with your own checks. A tool's completion message cannot tell you whether you selected every piece of information that needed removing.

5. Check the downloaded PDF

Open the newly downloaded file in your PDF viewer. Confirm the filename so you do not accidentally inspect the original.

1. Look at every page at high zoom. Check the edges of the black marks, repeated details and any unmarked pictures. Make sure the text you want to keep is readable.
2. Search for each removed word or code. Also try selecting and copying from the marked area into a plain-text document.
3. Check document properties for identifying information. Look for attachments, comments and form fields if your viewer exposes them.
4. Review the remaining visible context and the filename before sharing only the checked copy.

The output stores even the words you kept as pictures, so ordinary text search and copying no longer work.

The test: what was removed

We compared the actual Starlight download with an intentionally unsafe copy that merely covered content with black shapes. All ten fictional target codes remained recoverable from the unsafe copy through its text, images or other stored information.

For the Starlight download, separate PDF inspection found no target codes in extracted text or decoded objects. We also checked the output images with text recognition and looked at all three rendered pages. No target code was exposed, and the sentences meant to stay readable remained readable.

The output had no remaining form fields, annotations, attachments or bookmarks. The original identifying document-property value was absent. New software and date metadata remained, so describing the file as "metadata-free" would be inaccurate.

A small yellow fragment of a comment icon remained visible. Its comment content and annotation object were gone.

The hidden text layer was built into our test file; we did not test an OCR engine that creates this layer. Our test did not cover encrypted PDFs, signatures, unusual fonts or large files. The result supports this specific trial; your own document still needs checking.

If something is missing or still exposed

If search misses visible information, mark it manually and inspect the export. If a mark leaves an edge exposed, return to the working copy, cover a larger area and export again. Do not patch the downloaded PDF with a drawing rectangle and assume the content has been removed.

If the output loses a feature the recipient needs, keep it private and choose a workflow that preserves that feature. If you cannot verify a confidential file, stop before sharing and ask the person responsible for its release to review it.

Sources: Starlight PDF redaction (opens in a new tab), Starlight privacy (opens in a new tab), Starlight terms (opens in a new tab), Apple Preview (opens in a new tab), LibreOffice redaction (opens in a new tab).