A PDF carries a second document about itself
Besides its pages, a PDF holds a small set of document properties — title, author, subject, keywords, the app it was written in, the software that produced the PDF, and when it was created and last saved. Many files also carry the same facts again in an XMP packet, a block of XML that some apps write and others read. None of it appears on the page. Most PDF readers show it only under File → Properties, which is why it gets forgotten.
Even a PDF nobody consciously tagged has some. We printed a plain HTML page to PDF in Chromium 151.0.7922.34 with no settings changed, and it recorded the page title as Riverside Station Report, the creator as Chromium, the producer as Skia/PDF m151 — which names the browser's major version — and the exact second it was printed.
The test
We took that Chromium-printed 4-page report and gave it the kind of properties an office app writes: an author, a draft title, a subject, keywords, the name of the writing app, created and modified dates from months earlier, and an XMP packet naming the author again. We also added one thing people forget far more often than properties — a reviewer's sticky-note comment on page 1. Every name and value is made up.
That file then went through the exact code behind Merge & Split PDF three ways — merged with a second PDF, pages 1-2 extracted, and split into single pages — and through the same PDF.js render and canvas encode that the PDF to Image converter uses. Each output was opened and read back.
Merge, extract and split: the properties were dropped
| Property | In the tagged original | After merge, extract or split |
|---|---|---|
Title Often a working title or file name, drafts and version numbers included | Riverside Station - Q3 maintenance (DRAFT v4) | gone |
Author Usually filled in from the name on the account that saved the file | Jordan Example | gone |
Subject Rarely seen, sometimes revealing | Internal - pump replacement costs | gone |
Keywords Free text the author or a template added | budget confidential pump 2 failure | gone |
Creator (app) The application the document was written in | ExampleWriter 7.2 (Windows) | Replaced: pdf-lib (https://github.com/Hopding/pdf-lib) |
Producer The software that wrote the PDF, often with a version | Skia/PDF m151 | Replaced: pdf-lib (https://github.com/Hopding/pdf-lib) |
Created When the PDF was first made | 2026-05-11 08:42:00 UTC | Replaced: 2026-11-08 11:07:23 UTC |
Modified When it was last saved | 2026-06-02 17:15:30 UTC | Replaced: 2026-11-08 11:07:23 UTC |
XMP packet A second copy of author and app, in XML | present, names Jordan Example | gone |
The comment survived every time
This is the result that matters most. Comments, highlights and sticky notes are annotations, and annotations are attached to the page they sit on. Copying the page copies them.
| Output | Comment author | Comment text |
|---|---|---|
| Merged with a second PDF | Sam Reviewer | “Remove the supplier quote before this goes outside the company.” |
| Pages 1-2 extracted | Sam Reviewer | “Remove the supplier quote before this goes outside the company.” |
| Split, first page | Sam Reviewer | “Remove the supplier quote before this goes outside the company.” |
Removing the author from the properties while leaving a reviewer's name and an internal remark on page 1 is the kind of half-clean file that causes real embarrassment. Delete comments in the app that made the PDF, or turn the pages into images (below), before the file leaves your hands.
A black box over text is not a redaction
We drew a filled black rectangle over the heading of page 1 — the way people often “black out” a name — and then extracted that page. The heading text “Riverside Station Report” was still extractable from the covered original, and still extractable from the extracted page. The box is drawn on top; the words underneath remain part of the page, so selecting and copying, or any text extraction tool, recovers them. Page tools copy what is on the page, box and hidden text alike.
Rendering a page to an image: pixels and nothing else
Page 1 of the tagged file was rendered at 1191×1685 pixels and encoded as PNG and as JPG, the way the converter does it. We then listed every block inside each image file.
| Output | What the file contains | Document properties or comment text |
|---|---|---|
| PNG | Chunks: IHDR, IDAT, IEND — the header, the pixel data and the end marker. No text, EXIF or XMP chunks. | gone |
| JPG | A JFIF header, an ICC colour profile, and the compressed image. No EXIF or XMP segment. | gone |
An image of a page has no properties to inherit, no annotation layer and no text layer — what you see is all there is. That also makes it the one output here that genuinely removes text hidden under a box, because only the visible pixels are kept. The cost is real: text is no longer selectable or searchable, and at 1191×1685 the file suits screens rather than print. The PDF to image guide covers that trade-off in numbers.
What this test does not cover
- What is printed on the page. Names in headers, footers, signatures, file paths in footnotes and tracked changes that were printed into the PDF are page content, and survive every page tool.
- Every way a PDF can hide data. PDFs can also hold attached files, form field values, JavaScript and leftover objects from earlier edits. We tested properties, XMP and comments — the common cases — not every structure the format allows.
- Encrypted PDFs. Password-protected files cannot be opened by these tools at all.
Before you send a PDF
- Open File → Properties in your PDF reader and look at Title and Author.
- Open the comments panel and delete any comments, highlights or notes you did not mean to send.
- Never hide text by drawing over it. Use a real redaction tool, or render the affected page to an image with the PDF to Image converter.
- To send only some pages, extract them — the new file drops the original's title, author and XMP — then check its comments panel again.
Download the tagged sample, open its properties and comments in any PDF reader, then extract a page with the tool and compare.
pdf-metadata-sample.pdf (30 KB)Sharing photos rather than documents? The EXIF and privacy guide runs the same kind of test on a photo's location and camera data, and the EXIF Viewer checks a photo before you share it.