Is my file uploaded?
No. But “trust us” isn’t worth much on a page like this, so here’s exactly how FileSherpa works and how to check it for yourself.
How your file stays private
Most file sites say “secure” and leave it there. Here’s the actual architecture — your file and your typed request take different paths, and only one of them can ever leave your device.
Your file — never leaves
- Your file
- your browser
- processed here
- you save the result
What you type — may leave, if we can't work it out locally
- What you type
- matched on your device first
- if still unclear: sent to an AI model
Your file never travels on either path. What may travel is your typed request and a few measurements — that it’s a PDF, that it’s 18.4 MB, that it has 32 pages. Never its name, never its contents, never the file itself.
And you don’t have to take that on trust: press F12, open the Network tab, and run any tool on this site. Your file isn’t in the list, because it never left.
The eight questions worth asking
- Is my file uploaded?
- No. Every tool on this site runs inside your browser, on your own device. There is no upload step, which is also why there’s no progress bar and no queue — the work happens where the file already is.
- Is it stored anywhere?
- No, because it never arrives anywhere to be stored. This is a different claim from “we delete your files after an hour” — that requires them to have been sent first. Close the tab and nothing remains, here or anywhere else.
- Who can access it?
- Only you. Not us, not a hosting provider, not a sub-processor. There is no copy to access.
- Do I need an account?
- No, and you can’t make one. There is no sign-up, no email field, and no login anywhere on the site.
- Is AI involved?
- Only if you use the “describe what you need” box on the homepage, and often not even then — your wording is matched against the tools on your device first, which handles most common phrasings at no cost and with nothing sent anywhere. Picking a tool yourself, or running a route, never involves AI at all.
- What exactly gets sent to the AI, then?
- Your typed sentence, plus a few measurements of the file: its type, its size in bytes, and — where relevant — its pixel dimensions, page count, standard page size name, and whether it has bleed, selectable text, open paths, or location data. That’s the complete list. Never the file, never its bytes, never its name, never anything extracted from its contents — and for location data specifically, only the fact that it’s present, never the coordinates themselves. In the code this is enforced in one place, so it can’t be widened by accident.
- Are analytics used?
- Only if you accept the cookie banner. Decline (or simply ignore it) and no analytics script loads at all — it’s off by default, not on-until-you-object. When it is on, it records which tools get used and whether they worked. It cannot carry a filename, file content, or anything you typed; the event types are defined so those fields don’t exist.
- Are cookies used?
- Two things are stored in your browser regardless: your accept/decline choice (otherwise “decline” couldn’t stick) and your light/dark preference. Neither is a tracking cookie and neither leaves your device. Analytics cookies are set only after you accept, and you can change your mind from “Cookie settings” in the footer.
Don’t trust it — check it
This takes about thirty seconds and needs nothing installed:
- Press F12 (or Cmd+Option+I on a Mac) and open the Network tab.
- Drop a file into any tool on this site and run it.
- Watch the list. Your file isn’t in it, because it never left your machine.
The honest caveat, which is what should make the rest believable: if you use the typed request box and your wording can’t be matched locally, you will see one request, to our routing endpoint. Open it and read it — it contains your sentence and those few measurements, and nothing else. That is the only network request FileSherpa ever makes with a file of yours on screen.
Where a server does get involved
Being straight about the exceptions, since a page of unqualified reassurance isn’t worth reading:
- Serving the site itself. Loading this page means a server saw your IP address, the same as with any website.
- The typed-request endpoint. It briefly checks your IP against a request limit to stop automated abuse — held in memory for about a minute, never written to a log or database by our code.
- The AI model, if your request reaches it. Currently OpenAI, based in the USA, under the EU Standard Contractual Clauses.
If a future tool ever genuinely needs server-side processing, it will say so on its own page, in place of the “never leaves your device” badge — that line won’t be blurred.
The formal version, with legal bases, retention periods and your GDPR rights, is the Privacy Policy. Questions to blaskosamuel1@gmail.com.