All tools

How We Test PDFroo Tools

Many PDFroo pages quote results such as “12 pages in, 11 pages out” or “bookmarks were not kept”. This page explains how those results are produced, so you can judge them and repeat them.

Sample files

We use purpose-made files with fictional content, never real customer or personal documents. For the October 2026 guide tests these included:

  • a 12-page A4 report with a cover, a landscape page, 12 bookmarks, an internal link on the contents page, and a page marked “confidential” so we can check it does not leak;
  • image-only “scanned” PDFs with an upside-down page, a blank back side and a mix of A4 and US Letter sizes;
  • short multi-file packets, including a page saved with a 90° rotation flag.

The script that generates these files and the browser test scripts are kept in the site’s source repository with the tests.

How a test runs

  1. An automated browser (Google Chrome, headless, on Linux, desktop viewport 1280×900 unless stated) opens the live tool page, the same page visitors use.
  2. It adds the sample files and uses the same buttons and fields described in the guide.
  3. It records every network request the page makes while the file is processed. For the tools covered in the guides, no request carried the file: the only requests were for the site’s own pages and code.
  4. It downloads the result and we inspect it with independent tools: pypdf (page count, rotation, bookmarks, links) and pdftotext from Poppler (text content and position).
  5. Screenshots in guides are taken from these runs and cropped to the tool area.

Separately, an automated suite runs every public tool end to end before changes go live, checking for example that the merged file has the right page count or that a protected file is really encrypted.

What a test result means, and what it does not

  • A result applies to the sample file, browser and date shown. Very large files, damaged PDFs or unusual PDF features may behave differently.
  • We test on one desktop browser for the quoted figures; phone layouts are checked separately for usability, not for every figure.
  • We do not measure or publish speed claims across devices.

Re-testing

We re-run tests when a tool changes and at least monthly for the main tools, following our maintenance routine. The “Last tested” date on a guide changes only when a test was actually run. Changes are listed in the changelog.