Text placed outside the page
Text positioned beyond the page boundary, so no viewer draws it and every extractor still reads it.
How the text is hidden
The text matrix and the current transformation matrix place a run at a coordinate. Put that coordinate outside the box the reader displays and the glyphs fall off the sheet — the operators run, the string is in the content stream, and nothing appears. No colour trick, no font trick, no flag; only arithmetic.
Why a model still reads it
Extraction walks the content stream and does not ask where a run landed. Position is exactly the property a text extractor discards.
What we do about it
invisibleReason is ordered by strength of evidence, so a run that is off-page AND inside a switched-off layer, in the invisible render mode, collapsed to no area, or fully transparent is reported under that stronger reason instead. Off-page itself outranks the font-size test, so a run that is both off the page and below the sub-readable size is reported here rather than as pdf.tiny-font.
How often it fires
1% of 400 real UK public-sector contract PDFs (Contracts Finder, 2019–2023) — Word, Adobe, Nitro, office copiers, measured 2026-09-25.
This is an alert-volume number and nothing else. It says how often the alarm sounds on documents as found — not how often it is right, and not whether what it found was harmless. Documents as found may themselves carry concealment. Read it against the population named above rather than as a property of documents in general.
Seen in the wild
A benchmark of 29,322 PDFs hiding payloads behind invisible render mode, tiny fonts, white and low-contrast text and off-page placement. We ran our engine over every file: 9,774 of 9,774 injected documents flagged, and 0 of 9,774 ordinary ones. Read it against what the corpus is — the benign half is synthetic one-page templates, which the authors state plainly and we repeat here.
Across 196,682 real résumés submitted to a live hiring platform, roughly 1% carried a prompt injection hidden by white-on-white text, 1pt fonts or text placed off the page — and more than 90% of those prompts contained no explicit instruction at all, so a phrase matcher catches under one in ten.
A detector for the same problem this engine addresses, evaluated across 3,402 PDF and HTML documents; the evaluation set holds 113 documents carrying genuine hidden prompts — CVs, preprints, theses, a book chapter and blog posts — concealed by matching the background colour, invisible render modes, tiny text, layering, off-page placement, zero-area clipping, zero opacity, malicious fonts and hidden visibility.
Check your own file
Three commands: a key, credit, a verdict.
Start with the API