What "Free" Online Converters Actually Do With Your File
Nobody pays for a PDF converter with money, so somebody pays with something else. This is not an exposé of any one company — it is a description of an economic model, the clauses that express it, and a way to decide when uploading is a perfectly reasonable trade.
The free tool is not the product
Running a conversion service costs real money: bandwidth in, CPU time, storage, support, and the engineering to keep it up. A site offering that for nothing is funding it somewhere. The usual sources are advertising on the surrounding pages, a paid tier that removes limits, registration that converts visitors into accounts, or — least visibly — value extracted from the files and the usage data themselves.
These are not equally concerning. Ads are annoying but legible. A paid tier is honest. The last category is the one worth thinking about, because it is the only one where the file itself is part of the exchange.
What plausibly happens in the seconds after you click upload
Nothing here is sinister in itself; it is simply what a server-side pipeline looks like, and every step is a place your file exists:
- the file crosses the network and is received by a server, usually behind a CDN or load balancer that logs the request;
- it is written to temporary storage so a worker process can pick it up;
- the worker converts it, often producing intermediate files and a preview thumbnail;
- the result is stored somewhere so your browser can download it — which is why "download again" links exist at all;
- a cleanup job deletes it later, on a schedule stated in the terms.
Each of those is a reasonable engineering decision. Together they mean that for some window — minutes, hours, sometimes days — a copy of your file exists on hardware you do not control, governed by a retention policy you probably have not read.
The clauses worth reading
You do not need to read all of it. Open the terms or privacy policy and search for these five words. Their presence or absence tells you most of what matters.
- "retain" / "retention" — how long the copy lives. A stated period in hours is a good sign; silence is not.
- "licence" or "license" — whether you grant rights over what you submit. A licence limited to performing the conversion is normal and necessary. A broader, perpetual one is worth noticing.
- "improve" / "develop" — wording like "to improve our services" or "for development purposes" is the phrasing under which broader reuse of submitted content is typically granted.
- "third-party" / "subprocessor" — who else touches the file. Conversion, OCR and storage are frequently outsourced, and those partners are named here.
- "aggregate" / "de-identified" — data that stops being associated with you. This is usually genuine, and it is also the least of your worries if the file itself was the sensitive part.
Read the current text, not a summary — including ours. These clauses get revised, and a blog post from two years ago is not evidence about today.
When uploading is genuinely fine
It is worth saying clearly: most of the time, uploading is fine. If the file is a photo from your holiday, a public datasheet, or a draft you wrote and intend to publish, the convenience of a server-side tool outweighs a risk that is close to zero. Paying attention to this is not a lifestyle; it is a judgement applied to specific files.
The judgement flips when the file contains something whose disclosure would harm you or someone else. Contracts and drafts. Scans of identity documents. Medical images. Anything with a customer list, salary or bank detail. Unreleased designs, where the leak is competitive rather than personal. And — the one people forget — anything containing data about other people, where your convenience is spending their privacy.
There is also a compliance dimension that has nothing to do with the tool's honesty. If you are under an NDA, a professional confidentiality duty, or a data protection regime with cross-border transfer rules, sending a document to a server in another jurisdiction may be a breach regardless of how carefully that server behaves. In those cases local processing is not a preference; it is the only option that keeps you inside the rules.
How to tell a genuinely local tool from one that claims to be
Marketing language is cheap, so use tests instead of adjectives:
- Disconnect and retry. Load the page, switch off the network, run the job on a throwaway file. A tool that finishes offline did not use a server. This single test settles it.
- Watch what the page requests. DevTools → Network, then run a job. A real upload appears as a request with a payload roughly the size of your file.
- Notice what the tool asks for. A signup wall before a local-only job is a hint. Local processing has no need for an account.
- Check where the libraries come from. If the core processing library loads from a third-party domain, that party is positioned to see what happens.
- Read the "no upload" claim for qualifiers. "Deleted after 24 hours" and "never uploaded" are entirely different statements that get conflated in marketing copy.
Tools that pass all five are rare, which is largely why this site exists. The reasoning behind building it that way — including what it costs — is in Why every tool here runs in your browser.
Frequently asked questions
Do free converters keep my files?
Many retain uploads for a period to enable re-downloads, abuse checks or debugging. Whether they do, and for how long, is stated in their terms or privacy policy — look for a retention period, usually expressed in hours or days. If no period is stated anywhere, assume the file may persist.
Can my uploaded file be used to train AI models?
Some services reserve rights of this kind in their terms. The wording to look for is a licence to use submitted content to develop or improve services, or references to machine learning. Read the current terms rather than relying on summaries, as these clauses change.
Is uploading ever acceptable?
Yes, often. If the file contains nothing you would mind being public — a personal photo, a public datasheet, a document you authored and plan to publish — the convenience is usually worth more than the marginal risk. The calculus changes for contracts, IDs, medical images, unreleased designs and anything containing other people's data.
How do I tell whether a tool is genuinely local?
Load the page, disconnect the network, then run the job on a test file. If it completes, no server was involved. You can also open DevTools and check whether any request carries a payload the size of your file.
Related guides
- Why Every Tool on This Site Runs in Your Browser — the local-first architecture, its costs, and the offline test
- How to Redact a PDF Properly Before You Share It — masking versus real removal, and how to verify the result
- Offline Background Removal: Why Local Beats Free APIs on Privacy — when a local model is better than a free API
- How to Convert DXF to PDF Without Uploading Your Drawing — what drawings reveal beyond the geometry