PDF privacy guide · based on reproducible testing
How to tell if a PDF tool uploads your documents
HTTPS or a “secure” label does not tell you where your PDF is processed. To distinguish browser-local processing from remote transfer, observe what happens to the file during a real operation.
The short answer
Combine two signals: select the document only after taking the browser offline, and inspect network traffic in an equivalent online run. If the operation completes offline and no PDF content appears in processing requests, that supports local processing for that flow. Large POST or PUT requests carrying the file show remote transfer.
HTTPS does not mean “local processing”
HTTPS encrypts communication between your browser and a server. It matters, but it does not tell you whether the document stays on your device.
Offline testing: use the right order
Load the tool first, take the browser offline, then select a synthetic document. If you choose the PDF before disconnecting, it may already have been uploaded.
What to look for in network traffic
Inspect large POST or PUT requests, multipart/form-data, application/pdf, upload endpoints and temporary storage. A blob: result is compatible with local generation, but it does not prove the whole flow by itself.
Do not confuse telemetry with PDF transfer
Analytics and performance signals can leave the page without carrying the document. A small statistics event is not the same as a hundreds-of-kilobytes request containing the PDF.
Use a canary PDF, not a private document
Test unfamiliar services with a synthetic PDF containing a unique marker. Search that canary in requests, bodies, metadata and responses so you can trace the file without exposing real data.
POST PUT multipart/form-data application/pdf blob:
What our reproducible study observed
On September 5, 2026 we tested Merge PDF only across five services, using two canary PDFs and a second run with the network disabled after preloading each tool. The results describe that operation and date.
| Service | Observed result · Merge PDF · 2026-09-05 |
|---|---|
| UMyPDF | No transfer observed; completed offline after preload |
| Smallpdf | Remote document transfer observed |
| iLovePDF | Remote document transfer observed |
| PDF24 Tools | Transfer and remote merge request observed |
| Adobe Acrobat Online | Remote document transfer observed |
Quick checklist
- Use a synthetic document with no personal data.
- Load the tool and wait until it has initialized.
- Disconnect before selecting the file.
- Check whether the operation completes and produces a valid result.
- Repeat online and inspect POST/PUT, size, content type and domains.
- Separate web telemetry from document transfer.
- Record date, browser, exact operation and limitations.
Three conclusions to avoid
- “It uses HTTPS, so the PDF is not uploaded.” HTTPS protects transport, not processing location.
- “It works offline, so it never uploads files.” That only supports the tested flow when selection happens after disconnecting.
- “I saw analytics, so the PDF was uploaded.” Telemetry and document transfer are different things.
What a test can and cannot prove
A reproducible test can show an observed transfer or local completion under defined conditions. It cannot prove every mode behaves the same forever.
If you simply want to work with the PDF
PDFPrivado’s 14 published tools were audited with synthetic documents: after each tool was preloaded, all 14 completed their flow with the network disabled. The online control observed no document transfer during processing. A fully offline cold start is not part of that claim.
Frequently asked questions
Can an HTTPS PDF website upload my file?
Yes. HTTPS encrypts the connection, but the file can still be sent to a server.
Does working without Internet prove everything is local?
It is strong evidence for the tested flow when the file is selected after disconnecting; it does not cover every mode.
What does blob: mean?
blob: points to an object available in the browser. Interpret it together with earlier upload and processing traffic.
Should I test with a real document?
Prefer not to. Use a synthetic PDF with a unique marker.