We do not receive your documents

Your files

Every tool on this site runs inside your browser. The document you choose is read into your browser's memory, processed there, and released when you leave the page. It is never transmitted to this site or to anyone else, and it is never stored. Steady PDF cannot return the document later because the site never receives it. You can confirm this in your browser's network panel, or by using a tool with the device offline.

What you type

The text you search for in the redaction tool, the markup you paste into the HTML converter and the password you choose in the protect tool are used only by code running on that page. None of them is ever placed in the address bar, in browser storage or in a network request. A password stays in that tab's form and working memory long enough to protect the requested file; it is not sent to us or stored by this site. Steady PDF cannot recover a password you lose, while anyone who guesses it may be able to open the file.

What this browser remembers

If you pin a tool, open one, or change a setting inside a tool's options panel, that is written to this browser's own storage under the single name steadypdf.local. It holds the tool names you pinned, the last few tool names you opened, your chosen settings, and — once you have opened the Pro door — which of its two prices this browser was shown; nothing else, and nothing about any document. It never leaves the device, it is not an account, no identifier is created and there is nothing here to sign in to. The Clear what this browser remembers button under the tool list removes all of it. The cookies page lists every stored name in full.

Analytics

Every page loads one measurement script, from our own analytics service at analytics.pagefabrica.com. We run that service ourselves; it is not Google Analytics and it is not shared with an advertising network.

It sets no cookie and stores no analytics identifier in this browser. Nothing is written here for measurement — no cookie, local-storage entry or session value — so there is no client-side analytics record to clear.

What it receives is the address of the page you are on, its title, the address you arrived from, your browser's language and your screen size. From the request itself it derives coarse statistics — country, browser, device type — and forms an anonymous session hash from your network address and browser so that several page views in a day are counted as one visit rather than several. The network address itself is not stored.

It also receives product events, and only these eight names: work_started when you submit a tool, work_completed when it produces a result, and the Deep OCR preview's six — pro_offer_clicked_a or pro_offer_clicked_b (which of two prices was shown when you opened the Pro door), pro_preview_started, deep_model_loaded, deep_work_completed and deep_unsupported. None carries input or result properties — each call is an event name and nothing else. The compiled bridge accepts those eight names, refuses every other event name, and exposes no properties argument, so this path cannot attach a file name, page count, search term, password or document content.

The Pro door shows one of two prices and records which one you saw when you opened it, and then whether the preview was started, whether the model loaded, whether a read finished and whether this device was refused — the six names above, nothing attached. That is the whole of what "Pro" measures during the preview, in which Deep OCR is free, asks for no card and creates no account.

We use this to see whether the tools work and whether people can finish what they came to do, under our legitimate interest in running and improving them. The tag honours your browser's Do Not Track setting: with it on, nothing is sent at all. You can also block the address above, or write to support to object.

Advertising

Advertising positions are defined on some pages, but every current ad-unit identifier is empty: those elements are hidden, no advertising library is requested and no advertising cookie is set. When advertising is switched on, it will be Google AdSense. Google and its partners may then use cookies, device identifiers, approximate location derived from your IP address and interaction data to select, deliver, secure and measure advertisements. Nothing from your documents is available to them, because nothing from your documents is available to this site.

Where consent is required — the EEA, the United Kingdom and Switzerland — Google’s certified consent platform asks before any advertising purpose that needs consent, and provides a way to reject or manage those choices. Refusing is exactly as easy as accepting, in the same place and the same style, and nothing advertising-related loads until you have answered. That platform runs only when advertising is switched on; in the current version none of it is loaded.

The Deep reader's model

If you open the Pro door on an OCR tool and start a Deep read, this browser downloads the Deep reader's model — GLM-OCR, about 660 MB — from Hugging Face's servers (huggingface.co and its content network), pinned to one published version and checked here against digests this site carries. Your document is not part of that request, which names only the model's files; Hugging Face sees what any file download shows, the request and your network address. The model is then stored in this browser's cache under the name steadypdf-deep-aea46198f09e (the name carries the model's version, so a new version never reads the old files) so the next Deep read does not download it again; the Clear what this browser remembers button under the tool list removes it, and so does clearing this site's data in the browser.

Server logs

Like every website, the host that serves these pages records ordinary request information such as the address requested, the time, the browser's user-agent string and the network address of the request. That is the transport of the page itself; it never includes your document, its name or its contents.

Your rights and your questions

Because we hold no account, no profile and no file, there is usually nothing for us to show you, correct or erase. You can still write to support with any question about this policy. Please do not attach the document you were working on: we do not need it, and we would rather not have it.