Redact PDF on Android

What redaction means here

A black rectangle drawn over a page covers; it does not remove. The characters underneath stay in the file, selectable by anyone who opens it in a plain reader. Redaction in PDFMason is the other thing: the text and image data under every mark are deleted from the file itself, so nothing under the marks can be selected, searched, copied or recovered from the result.

Deletion alone is not the whole job, either. A PDF carries more than what its pages show: document information, XMP metadata, embedded files and attachments, JavaScript, and earlier revisions left by incremental saves. Redacting the visible line but shipping the file with the same sentence still sitting in its metadata is the classic failure. The redaction pass sweeps those alongside the marks.

The side-by-side of the three approaches — painting over, cropping, and deleting — with the checks you can run on each, is in black boxes versus real redaction.

Two ways in: manual marks or rule-assisted search

Manual redaction works like a marker pen with consequences: draw boxes over anything you want gone, page by page. Use it when you already know where the content is — a signature, a table cell, a paragraph.

Auto-redact finds candidates for you when you do not know where everything is. It runs explicit rules on the phone: phone numbers and government ID numbers for the regions you pick, international bank account numbers, bank card numbers with checksum validation, email addresses, web addresses, and any custom text you add. Several regions can be scanned at once. The rules are a search aid with a defined reach — they find what the rules describe, not everything sensitive in the document.

• Candidates are listed per page with context, the rule that matched, and a highlighted preview.
• You tick what goes: nothing is marked or deleted until you confirm.
• A match whose position cannot be pinned reliably is counted but never marked on its own.
• Scanned pages are recognized with on-device OCR first, so rules can match inside images of text.

Both entrances run the same pipeline underneath — the difference is only who finds the content, you or the rules.

Before anything is written

Confirming the marks produces a summary, not a silent write: how many marks merge into how many regions, which pages are affected, which ranges must grow to cover an object that cannot be split, and — the part most tools leave out — anything that cannot be deleted safely, each with its own reason. A gradient that cannot be partly deleted, a mark covering only part of a line, a shape that cannot be split, an image that cannot be read: each says which it is and what to do about it. Nothing is written while a blocker stands.

Limits, stated plainly

The product-wide limitation list, in the same words used everywhere on this site:

Password-protected files, browser versus app: “The browser tools can't open password-protected PDFs. The Android app can.”

The browser side is deliberately smaller than the app: “No OCR, no compression, no PDF-to-Word in the browser — yet. We'd rather ship nothing than ship a version that doesn't work.”

Line and arrow annotation shapes were tested on real documents and cut: “We tested it on real documents: it only writes correctly to 33% of them. So it's not in v1.”

Saving very large documents: “A 1,136-page document takes 24–32 seconds to save with verification. It runs in the background and you can cancel it.”

Compare works on the text layer, not the pixels: “Images, formatting-only changes and paragraphs that merely moved are not compared — two files that look different but read the same may show no differences.”

Find and replace refuses what it cannot fit: “A replacement that would overflow its frame is flagged and left unchanged rather than silently shrunk — you decide per match, and scanned or low-confidence pages are listed as out of reach instead of edited blindly.”

Camera scanning depends on the device: “Camera scanning uses Google’s document scanner, so it appears only on devices with Google Play services, a camera and enough free memory; everywhere else the entry is hidden and photos can be imported instead.”

Repair is best-effort by design: “Repair rebuilds damaged pages but drops interactive forms, annotations, links, bookmarks, signatures and attachments, and cannot promise that missing content comes back.”

Specific to redaction: the rules find what they describe — chosen regions, formats, and your custom terms — and are not a guarantee that every sensitive item in a document has been found. A person still reads the list. Marks are boxes for now; selecting a phrase and marking it directly is not yet a gesture the app supports. Very large jobs have a stated ceiling of ten thousand matches.

To remove a whole file’s metadata without redacting anything, Document properties in the app clears it in one place. Password protection for the finished copy is on protecting PDFs. The app is on Google Play.

FAQ