Redact PDF
Permanently remove sensitive content from a PDF — true removal, not just a black box on top; the rest of the page stays real text
Want AI to do the finding? Try AI PDF Redactor (sign-in required).
Drop a PDF fileTap to choose a PDF file
or click to browseor tap to choose
Mark and permanently remove sensitive content
Large files (over 50 MB) may take longer or run out of browser memory
Watch: Redact PDF in Action
Permanently removes sensitive information from a PDF. The text, picture pixels, drawings and form fields under each box you mark are taken out of the file itself, not just covered, and a black bar is drawn in their place. Everything else on the page stays as it was: the rest of the text is still real, selectable and searchable, in its own fonts. A simple black box drawn on top of text in most "redaction" tools deletes nothing, because the real text is still in the file for anyone who copies it or opens the PDF's raw content. This tool removes what is underneath.
A page is turned into a picture only when it cannot be edited exactly: a curved drawing partly under a box, a shading, an unusual font encoding, or a picture format the tool cannot read. When that happens the page's text can no longer be selected, searched or read by a screen reader, and a notice under the result names the page and the reason; run OCR PDF on the result if you need searchable text on it again. Every other page keeps its real text, links and bookmarks. Search reads the page text, including a name split across text pieces, and boxes only the matched words. It can't read scanned pages (they're pictures) — draw boxes by hand or run OCR PDF first. With the option on, link addresses, form values and bookmarks that contain the searched text are removed, and every comment-style annotation on unmarked pages is removed whether or not it mentions the text. It's still worth looking over the result before you send it. What is and is not cleaned: only the areas you mark (or that a search, detector or custom pattern marked and you kept) are removed from the page. Document properties and the XMP metadata are cleared only when "clear document metadata" is on. Comments are all removed when "remove comments..." is on, but links, form values and bookmarks only when they contain the searched text, a ticked detector's match or a custom pattern's match. Attached files (and what is inside them) are never inspected or removed, and the summary counts any that remain. Custom patterns are checked before they run: at most 5 patterns, 60 characters for a simple pattern and 100 for a regular expression; lookahead, back-references and ^ $ are not supported, and a pattern that would be too slow is stopped with a message instead of freezing the page.
- Can I still select and search the text on a redacted page?
- Yes. Only what lies under your boxes is removed; the rest of the page keeps its real text, fonts, links and bookmarks. The one exception is a page that cannot be edited exactly, which is turned into a picture of itself with the boxes painted black. The notice under the result names any such page and why, and its text is then no longer selectable or searchable. Run OCR PDF on the result if you need it searchable again.
- When is a page turned into a picture, and how will I know?
- Only when the content under a box cannot be cut out exactly: a curved or irregular drawing that sticks out of the box, a gradient fill, an unusual font encoding or writing direction, a picture in a format the tool cannot read, or a picture with a stencil mask. The result then lists each such page with its reason in a red notice, and the message under the Download button says how many pages were affected. If no notice appears, no page was turned into a picture.
- What does Select all detectors do?
- It ticks every automatic detector at once (email addresses, phone numbers, card numbers, SSN-style numbers, EIN-style numbers and likely names). Click it again to clear them. Likely names is a best guess, so review every mark before you apply.
- Does redacting actually remove the content, or just hide it visually?
- It permanently removes the underlying content in the redacted area, not just draws a black box over it visually — an important distinction, since a visual-only cover can sometimes still be selected or extracted from the original data underneath.
- Can I search for text instead of manually drawing every box?
- Yes — type what you're looking for (a name, account number, or any other text) and click Find & Mark to automatically mark every place it appears across the document. Review the marks before applying — search finds exact matches within individual text runs, so a name split across two separate runs (an unusual line wrap, for instance) may not be caught, and manual marking is still worth double-checking for anything critical.
- What's the difference between this and Watermark?
- This genuinely removes content from the document. Watermark is purely visual — it stamps text or a logo on top of the page without removing or hiding anything underneath, and the original content stays fully present and extractable. If the goal is actually deleting sensitive information, use Redact; if the goal is labeling a document (draft, confidential, branded), use Watermark.
- Does this also remove hidden metadata, like the document's author name?
- Only if you check "Also clear document metadata" before applying — that clears title, author, subject, keywords, and producer/creator fields, since a PDF's metadata can independently reveal who created it even after the visible content is redacted. It's off by default since not every redaction needs it.
- Will pages I don't mark be affected at all?
- No — pages with no redaction marks are not touched. On a marked page only the content under your boxes changes; a page is turned into a picture only if it cannot be edited exactly, and the tool says so.
- Are my files uploaded to a server?
- No — both marking and applying redactions, including text search, run entirely in your browser. Your PDF never leaves your device.
- Can I undo a redaction after downloading?
- No — once applied and downloaded, the redacted content is genuinely gone from that file, not recoverable. Keep your original, unredacted file if you might need the removed content again for a different purpose.
- What should I do after redacting if the document needs to stay restricted?
- Use PDF Password next to require a password to open the redacted file — redaction removes specific content, but doesn't by itself restrict who can open the resulting document.
- Is the text under a black box really deleted?
- Yes. Every character whose box touches a redaction box is deleted from the page's drawing instructions, including invisible text, and the letters after it keep their place. A picture that overlaps a box has the pixels under the box painted black and is stored again, so the original pixels are not in the file; a picture is re-saved in the same way even where only a corner of it is covered, so a JPEG there is re-encoded. Lines and rectangles are cut to what lies outside the box, and links, notes and form fields under a box are removed along with their values. Page thumbnails and the alternative-text labels of a tagged PDF are dropped from marked pages, and the old versions of changed content are deleted from the file before it is saved. Text on a page you did not mark is not touched, so mark every page that needs it.
- What do the comment and metadata checkboxes remove?
- The first deletes every comment-style annotation on every page, plus links and form values that contain your searched text and bookmarks whose titles match; the outline is rewritten only if a bookmark had to go. Clearing metadata empties every entry of the document information (title, author, subject, keywords, creator and any custom ones) and drops the XMP metadata.
- What do the pattern checkboxes match?
- Email addresses, ten-digit phone numbers with an optional country code, card-style runs of 13 to 19 digits with optional spaces or dashes, and SSN-style numbers written like 123-45-6789. They match formats only, so other long digit runs can be marked too; review the boxes.
- Why does search find nothing in my scanned PDF?
- Search reads the page's text layer, and a scan is just a picture. If no page has text, the tool says so and suggests drawing boxes by hand or running OCR PDF first. In a mixed file only the pages with text can be searched.
- What does an empty marked area warning mean?
- When none of a page's boxes had any text, picture, drawing or form field under it, the tool warns that nothing was found under the marked areas on that page, a hint that they may sit in the wrong place. The black bar is still drawn. On a page that became a picture the check looks for non-white pixels and names the number of empty boxes instead.
How to use Redact PDF
There are two ways to mark content. To search, type a name or number in the box, one term per line, optionally tick Match case, Whole words only, and the patterns Email addresses, Phone numbers, Card numbers or SSN-style numbers, then press Find & Mark. Matches appear as black boxes on the thumbnails with a red count badge. To mark by hand, click a page, drag a rectangle over the area, and hover a box to reach its X.
Choose whether to remove comments, links, form values and bookmarks that contain the searched text, and whether to clear metadata. Press Apply Redactions, read any warnings, check the preview, and download the file named redacted_ followed by your original file name.
When to use it
Before sharing a contract, case file or bank statement in which names, account numbers, home addresses or signature images must go. The pattern checkboxes help when a document is full of email addresses or phone numbers, and the multi-line search suits a list of names that must vanish everywhere they occur.
When not to use it
Scanned pages hold no text, so Find & Mark finds nothing there; draw boxes by hand or run OCR PDF first. Marked pages are edited in place, so the rest of their text stays searchable; only a page that cannot be edited exactly becomes a picture, which this tool reports. To only label a document use PDF Watermark, and to lock the whole file use PDF Password.
Tips for best results
Search first, then scroll through every thumbnail, because matching depends on the text layer: the search joins words split across lines and lets a space in a term match any run of whitespace, but unusual spacing can still be missed. Add alternative spellings on separate lines. Boxes reach a little above and below the letters to cover tails. Afterwards, try selecting text near a box on a redacted page: the rest should still select, the covered words should not, and heed any warning about a page that became a picture or a box with nothing under it.
Limitations
On a marked page, the text characters, picture pixels, line and rectangle drawings and annotations under the boxes are removed from the page's own drawing instructions and a black bar is drawn; the rest of the page keeps its real text and fonts. A picture that overlaps a box is decoded, painted black under the box and saved again, so a JPEG there is re-encoded. Where the content cannot be cut exactly (a curved shape sticking out of a box, a shading, an unusual font encoding, a picture format the tool cannot read) that page is rendered to a JPEG at three times its size with the boxes painted in and replaces the page; the notice under the result names those pages. Pages without marks are not touched, so sensitive text on them remains.
Privacy
Editing the pages, and drawing any page that has to become a picture, happen inside your tab, and the code that does it is loaded only when you open this tool. The original is only read, never changed, and the redacted copy exists in memory until you download it. Searching uses the PDF reader bundled with the page, not a remote service, and your search terms stay in the page.
More from PDF Tools: