People share PDFs as if they were printed pages. A PDF is closer to a container. The visible page is one part, and the file can also hold the name of the author, the program used, comments from reviewers, files attached inside it and text that is covered but not deleted. When a report, CV or court filing goes out, those extras go with it.
This guide lists the hidden information a PDF can carry, explains why highlighting or a black shape is not removal, and ends with a checklist to run before sending. The step-by-step method for removing covered text is in how to redact a PDF for real, and is not repeated here.
What official guidance says about hidden data
Government bodies have warned about this for years. A 2005 guide from the US National Security Agency's Information Assurance Directorate, "Redacting with Confidence", warns that converting a Word file to PDF "does not remove all metadata automatically". On covering text, it says the method is "not effective, in general, for computer documents distributed across computer networks", and that "information hidden or covered in a computer document can almost always be recovered". The guide is old and was written for Word and Acrobat of that period, but the principle has not changed.
The UK Information Commissioner's Office gives similar advice for organisations releasing information. Its disclosure guidance says that files "rarely contain just the information entered by the author or just what is displayed on the screen", that metadata can include previous authors and changes to earlier versions, and that text hidden by matching its colour to the background "will still remain within the source file". On black highlighting it warns that "the information still exists underneath the black box in the original electronic file". It suggests final checks such as "Are all redactions effectively applied?" and "Is there any meta-data that should be removed?"
What a PDF can carry
Not every PDF holds all of these. Each is a place to look.
- Document properties. Title, author, subject, keywords, creation and modification dates, and the program that made the file. A CV exported from a work computer can show the employer's software and the author's username.
- XMP metadata. A second, richer block of metadata in XML form, often repeating the same details and sometimes adding more, such as the edit history of the source file.
- Comments and annotations. Sticky notes, text comments, pencil marks, replies from reviewers and highlights. These are objects stored on the page, separate from its text, and can show names and dates.
- Form data. Filled-in form fields keep their values in the file, including fields that sit off the visible area or have been hidden.
- Attachments. A PDF can contain whole files inside it, such as the original spreadsheet or Word document.
- Hidden layers. A PDF can have layers that are switched off by default. A layer that is not shown is still stored.
- Text under shapes. Text covered by a black rectangle, a white box or a highlight remains as text in the page. It can be selected, copied or extracted.
- Earlier content. Some programs save changes by adding to the end of a file, which can leave older content in place. This depends on the program that saved the file, so it is worth checking and not assuming.
The PDF format itself is an open ISO standard, ISO 32000, and the PDF Association provides the ISO 32000-2 specification free of charge.
Why highlighting or a black shape is not removal
A PDF page is a list of drawing instructions. A line of text is one instruction. A highlight or a black rectangle is another instruction drawn after it, or an annotation placed above the page. Adding a shape adds an item. It does not delete the text instruction.
Highlighting is the same. A yellow highlight changes what the eye sees, and a black highlight hides the words visually, but both leave the characters in the file. The ICO guidance describes exactly this: saving or printing to PDF is "unlikely to provide effective redaction" because the highlighting formatting can be kept, and selecting the highlighted text and pasting it into an editor can expose the content.
The reliable alternatives are to remove the text from the page, or to replace the page with an image in which the covered area is black pixels. The ICO describes dedicated redaction software that converts the PDF to an image so the text "cannot be copied", with the trade-off that it is no longer machine-readable. The UK National Archives redaction toolkit recommends the bitmap route too.
Common mistakes
- Covering a number with a black box or white rectangle and sending the file.
- Marking text with the highlighter and assuming it is hidden.
- Sending a PDF without opening its properties, so the author's name and software go out.
- Leaving reviewer comments in the file because they are collapsed.
- Redacting a copy but sharing the original by mistake. Name the safe copy clearly.
- Hiding information in white text or an off-page area. The text is still stored.
- Not testing the result. Searching the finished file for the removed words takes seconds.
Checklist before sending a PDF
- Work on a copy and keep the original.
- Open the document properties and read the author, title, dates and software.
- Look for comments, notes and highlights, and decide whether each should stay.
- Check the attachments or paperclip panel in a PDF reader for embedded files.
- Check the layers panel, if the reader has one, for layers that are switched off.
- For any covered text, confirm the text was removed and not just covered. Try to select it, and search for it.
- Remove the metadata from the final file.
- Open the finished file in a different PDF reader and run the searches again.
- Send the finished file, and not the working copy.
How to do it with DocPro
Two free DocPro tools cover the metadata and the covered text. They run in the browser and the file is not uploaded. No account is needed.
Cleaning metadata. Open Clean PDF metadata and choose the PDF. The tool lists what it finds under "Document information" and "Other traces", which show XMP metadata, the document ID and hidden application data.
- Review the fields shown. They can include the author, title, dates and the software that produced the file.
- Under "Remove", leave ticked what should go: "Document information (author, title, dates, software)", "XMP metadata", "Document ID" and "Hidden application data". Tick "Keep the title" if the title should stay.
- Press "Clean metadata" and download the copy. The tool reopens the result and reports whether the selected items are gone.
Removing covered text. Open Redact PDF and choose the PDF. Under "Draw boxes", drag over each area, or type a word into "Word, name or number to find" and press "Find and box all". Check every box. Press "Redact and save PDF". Every page with a box becomes an image with the boxes painted black, and the tool checks that no selectable text is left on those pages. The option "Remove document metadata (author, title, dates and software)" is ticked by default.
The result is a new file built from scratch, so document-level extras from the source file are not copied into it. The result card notes that bookmarks, attachments and form fields are not carried over.
What DocPro does not remove
- Clean PDF metadata removes document information, XMP metadata, the document ID and hidden application data. It does not remove comments, annotations, attachments, form values, hidden layers or text on the page. Its own tips say that metadata is separate from the visible pages and that Redact PDF is the tool for text on them.
- Redact PDF keeps pages without boxes as they were, apart from application data and metadata links on those pages. Comments or other objects on an unredacted page may therefore stay. Check those pages, or box a small area on them so the page is rebuilt as an image.
- Redacted pages become images. The text on them can no longer be selected or searched, and the file can be larger. If readable text is needed again, OCR PDF can add it back, but only for the parts that remain visible.
- Neither tool is a complete sanitizer. Acrobat Pro includes a Sanitize function that Adobe's help describes as a way to "remove sensitive and hidden information from your PDFs". For a legal or regulated release, use a full sanitizing tool, test the output, and follow the rules of the receiving body.
- Flatten turns annotations and form fields into fixed page content, which stops them being edited. It is not a substitute for deleting a comment that should not be shared.
A last check is simple. Open the finished file, press Ctrl+F and search for a word that should be gone. Then try to select the area that was covered. If anything is found, the file is not ready to send.
Frequently asked questions
Does highlighting text in a PDF hide it?
No. A highlight is an annotation drawn over the text. The characters stay in the file and can be selected, copied or extracted. To remove text, delete it or replace the page with an image where the area is black.
Can text under a black box be recovered?
Yes, if the box is only a shape drawn on the page. Official guidance from the NSA and the UK ICO states that information covered in a computer document can almost always be recovered unless it was removed.
What metadata does a PDF contain?
Commonly the title, author, subject, keywords, creation and modification dates and the software that made it. A PDF may also hold XMP metadata, a document ID and application data. Open the document properties in a PDF reader to see them.
Does DocPro remove comments and attachments from a PDF?
Clean PDF metadata does not. It removes document information, XMP, the document ID and hidden application data. Redact PDF builds a new file where pages with boxes become images, and its result card notes that bookmarks, attachments and form fields are not carried over.
Is DocPro's metadata cleaner free?
Yes. Clean PDF metadata and Redact PDF are free, run in the browser, and need no account. Only Sign PDF and saved signatures are Pro.