1. Document properties
Title, author, subject, keywords, creator and producer travel with every PDF and are filled in automatically — the author is usually the login name of whoever created the template. Open Remove metadata, read what the file says, correct the title, and strip the rest. Do this for every document built on someone else’s template.
2. Text under black boxes
A rectangle drawn over text does not remove the text; select-all and copy recovers it, and this is how redaction failures make the news. Redact re-renders the marked pages as flat images with the boxes burned in, so there is no text layer under them. After redacting, open the result and try to select text on those pages — there should be none. The redaction guide covers the details, including pattern and on-device AI suggestions for finding what to box.
3. Comments, notes and links
Reviewer comments are annotations that sit above the page and are invisible in many previews. A client can open them in Acrobat. Flatten with annotation removal strips them; Comment & link lists existing links so a tracking URL or a link to an internal server can be removed.
4. Form data
A filled form keeps its fields editable, which means answers can be changed after you send it and, in some viewers, the field history is visible. Flatten after filling so the answers become page content. If the fields must stay live for the recipient, know that you are sending an editable document.
5. Revision history and deleted pages
Editors that save “incremental updates” append changes and keep old objects, so a PDF can contain earlier versions of itself, including pages you deleted. Rewriting the file from scratch removes them: Compress on the Print target is lossless and does exactly that. Delete pages also rebuilds the file from the kept pages only.
6. Embedded files and hidden content
PDFs can carry attachments, layers and scripts. The PDF/A preflight lists embedded files, JavaScript and multimedia, and conversion removes them — useful even when you do not need PDF/A. Extract images shows every image object, including ones covered by other content.
7. The channel
An email attachment lives in two inboxes indefinitely and passes through servers you do not control. For personal, medical, legal or financial content, either encrypt with a password sent by a different channel, or use Send securely: the file is sealed in your browser with an AES-256 key, only ciphertext is uploaded, the key travels in the link fragment that browsers never send to servers, and the link expires after a time and a download count you choose. For proof that the file has not been altered since you sent it, certify it; the recipient can verify on their device.
Scans and photos have their own leaks
A scanned page carries no hidden text layer — until you run OCR, at which point every word on it becomes searchable, including the ones you meant to cover. Redact after OCR, not before, and check the redacted pages for selectable text. Phone photos embed EXIF data: the camera model, the date and often the GPS position where the picture was taken. Converting them with JPG to PDF keeps the image bytes as they are, so strip location data in your photo app first, or use Scan from phone, which sends straightened page images rather than the original photos.
The order
- Delete pages, redact content.
- Fill and sign; flatten.
- Strip metadata; set a title.
- Rewrite (Compress on Print) to drop history.
- Certify if integrity matters.
- Protect or seal, last — most tools cannot operate on an encrypted file.
Every step runs in your browser. The document you are cleaning is precisely the one that should not be uploaded to clean it; the security page sets out what leaves your device and what never does.