INDEPENDENTLY REPEATABLE CHECKS
PDF tool privacy tests: what each test actually proves
“No upload” is a technical claim. A useful privacy review combines several tests because no single check answers every question.
| Test | What it proves | What it does not prove |
|---|---|---|
| Offline processing | Processing works after disconnecting | Whether data was sent before or after the offline job |
| Network inspection | What the page sends during the test | Future behavior or hidden requests outside the captured session |
| Browser storage review | What remains in local/session storage and IndexedDB | What exists only in memory or was already sent |
| Output metadata review | Whether exported files expose author, software or timestamps | Whether the input was uploaded |
| Privacy-policy review | What the provider promises and which services are named | The site's actual runtime behavior |
A reproducible privacy test protocol
- Use a synthetic PDF containing a unique marker, never a confidential document.
- Record browser version, tool URL, date and test-file checksum.
- Capture the Network panel while opening, processing and downloading.
- Repeat offline and record whether the same operation completes.
- Review browser storage before and after closing the tab.
- Inspect output metadata and publish both successful and failed observations.
Privacy expectations by PDF job
- Merge and split: no document bytes should leave the browser.
- OCR: extracted text is sensitive and should remain local too.
- Redaction: removed content must not remain recoverable in the output.
- Metadata removal: verify the exported file, not only the interface message.
- Conversion: confirm whether third-party fonts, APIs or remote renderers are used.
Follow the full walkthrough: How to verify a PDF tool does not upload your files →