PDF Hidden Content Finder
Cropping a PDF page does not delete anything. It draws a smaller window over the same content, and every word you pushed outside that window is still in the file — one select-all, one layout tool, one changed display setting away. Drop a document here to see what the person receiving it can get back.
🔒 Your document never leaves your browser. The crop is undone in memory, read, and thrown away — nothing is uploaded and nothing is written back to your file.
Drop the PDF you are about to send
One PDF at a time · nothing is uploaded
Why a text extractor will not show you this
This is the part that makes cropped content dangerous rather than merely careless. When you open a cropped PDF in a reader — or run it through a text extractor, including our own — the library reports the page as the crop box and returns only the text inside it. The cropped words are not marked as hidden and they are not returned as empty strings. They are simply absent, and the output looks complete.
So the usual way of checking a document before sending it — extract the text, read it, decide it is fine — cannot find this class of leak at all. The check has to undo the crop first. That is what this page does: each page’s crop box is pushed back out to the full media box in memory, the document is read again, and the text that falls outside the original crop is pulled out and shown to you as text.
Where cropped-in content comes from
Almost never from someone trying to be devious. It comes from a design file that had the internal margin notes on the pasteboard, from a report where the sensitive column was cropped off rather than deleted, from a scanned contract trimmed to hide a signature block, from a quote where last quarter’s pricing sits just past the page edge. In each case the person who cropped it saw the result on screen, saw the content gone, and reasonably concluded it was gone.
The fix is always the same and it is not a setting: delete the content, or export a flattened copy where the area outside the crop genuinely no longer exists — then check the result here again.
What this tool cannot see
Stated plainly, because a clean result has to mean something. The judgement covers the text layer only. Images and vector art outside the crop are recoverable in exactly the same way and are not visible to this check — seeing them would mean rendering every page at full media size, which is a different tool. A page that is cropped with no text finding is reported as such (rule REC-002) rather than as “nothing is hidden”.
Layers that are off by default are listed by name; the content drawn inside them is not extracted, which would need the content stream split by its optional-content marks. Incremental-save rounds are counted, not reconstructed. And black boxes drawn over text are a different failure entirely — that one belongs to the redaction checker, which looks under the rectangles while this page looks past the page edge.
Every check this tool runs (6 rules)
REC-001 There is text outside the crop box, and here it is
Cropping a PDF page does not delete anything. It sets a CropBox, and readers draw only what falls inside it — every glyph outside stays in the file and comes back the moment someone selects the text, opens the file in a layout tool, or switches their reader to show the MediaBox. This is the check the tool exists for: the words outside the crop are pulled out and shown, because a count of "hidden objects" does not tell you whether what leaked matters.
Fix: Do not crop to hide. Delete the content, or export a flattened copy where the cropped area is genuinely gone, then re-check.
Source: yaktool engineering rule, not a citation: PDF 32000 defines CropBox, optional content and incremental updates; treating each of them as "content the recipient can still recover" is our reading, and the tool says which findings are judgements rather than facts of the format.
REC-002 The page is cropped, so images may extend past it too
The CropBox is smaller than the MediaBox, which means part of every object on the page lies outside what a reader shows. No text was found out there, but images and vector art are cropped the same non-destructive way and this tool judges the text layer only — seeing the pixels would need the page rendered. Reported so the absence of a text finding is not read as "nothing is hidden".
Source: yaktool engineering rule, not a citation: PDF 32000 defines CropBox, optional content and incremental updates; treating each of them as "content the recipient can still recover" is our reading, and the tool says which findings are judgements rather than facts of the format.
REC-003 The file has layers that are switched off by default
Optional content groups are the PDF equivalent of layers, and one that is off when the document opens is still fully present — a reader with a layers panel turns it back on. They are usually left over from the design file: an earlier price, an internal annotation, a watermark someone decided against. The layer names are listed; what is drawn inside them is not extracted here, which would need the content stream split by its optional-content marks.
Fix: Flatten the document, or delete the layers you do not intend to ship.
Source: yaktool engineering rule, not a citation: PDF 32000 defines CropBox, optional content and incremental updates; treating each of them as "content the recipient can still recover" is our reading, and the tool says which findings are judgements rather than facts of the format.
REC-004 The document properties carry names, software and file paths
Author, company, document title, and sometimes the full path of the original file on somebody’s machine travel in the metadata and never appear on any page. The values are shown rather than counted, because "3 metadata fields removed" does not tell you that one of them was a client name. Producer, creator and timestamps are listed alongside them but do not raise this finding on their own — every PDF has them, and a warning that fires on every file is a warning nobody reads.
Fix: Clear the document properties before sending, and check again — some tools write them back on every save.
Source: yaktool engineering rule, not a citation: PDF 32000 defines CropBox, optional content and incremental updates; treating each of them as "content the recipient can still recover" is our reading, and the tool says which findings are judgements rather than facts of the format.
REC-005 The file keeps its earlier versions
Saving incrementally appends the changes and leaves everything before them in place, so a document edited and re-saved several times contains its own history. This tool reports how many save rounds are present; it does not reconstruct the earlier text, but anyone with a PDF parser can walk back through them.
Fix: Save a fresh copy (or "save as", not "save") so the file is rewritten without its history.
Source: yaktool engineering rule, not a citation: PDF 32000 defines CropBox, optional content and incremental updates; treating each of them as "content the recipient can still recover" is our reading, and the tool says which findings are judgements rather than facts of the format.
REC-006 Files are travelling inside this document
Attachments ride along inside a PDF without appearing on any page. Often that is the point — a hybrid invoice carries its XML this way — and sometimes it is a spreadsheet somebody dragged in and forgot.
Source: yaktool engineering rule, not a citation: PDF 32000 defines CropBox, optional content and incremental updates; treating each of them as "content the recipient can still recover" is our reading, and the tool says which findings are judgements rather than facts of the format.
Thresholds are published, not hidden: a text run counts as outside the crop when its centre point falls outside the crop box, and a page counts as cropped when any edge of its crop box is more than 0.5 pt inside the media box — the tolerance exists because many producers write the crop box back as a float re-read of the media box.