Which web tools keep working offline
If a web tool still works after you disconnect from the internet, it cannot be uploading your files. That makes the offline test the simplest privacy check available.
Ahmed Kobbi · · 5 min read
How can a web page work without the web?
It sounds contradictory, but the explanation is straightforward. A web page has two phases: loading and running. During loading, your browser fetches the HTML, CSS and JavaScript that make up the page. Once that has finished, the JavaScript is running on your machine, using your processor and your memory. Whether the page needs the network after that point depends entirely on how it was built. A page that sends your data somewhere for processing needs a connection every time you use it. A page that does the work in JavaScript needs the network only once, to load itself. Modern browsers are far more capable than most people assume. They can parse and write PDFs, decode and re-encode images, compute cryptographic hashes, generate secure random numbers and manipulate multi-megabyte files entirely in a tab. Ten years ago much of this genuinely required a server. It no longer does.
The offline test
This gives you a privacy check that requires no technical knowledge and cannot be faked. Load the tool's page normally. Then disconnect: turn off Wi-Fi, unplug the ethernet cable, or switch on aeroplane mode. Now use the tool — drop in your file, run the operation, download the result. If it works, the processing happened on your device. There is no alternative explanation. A tool that uploads your file cannot function without a network connection, and no amount of marketing language changes that. If it fails, hangs, or shows an error, the tool needs a server. That does not automatically make it untrustworthy, but it does mean your file is being transmitted somewhere, and you are relying on a policy rather than on architecture. The test is more reliable than reading a privacy policy, because it observes behaviour rather than intent.
What genuinely cannot work offline
Being honest about the boundary matters, because a tool that claims everything works locally when it obviously cannot is a warning sign. Anything that fetches external data needs a network by definition. A YouTube thumbnail downloader has to request the image from YouTube's servers — the URL parsing happens locally, but the image itself must be fetched. Live currency conversion needs current rates. Anything checking a domain, an IP address or a certificate needs to reach it. Heavy machine-learning work is a grey area that is shifting. Background removal, upscaling and OCR traditionally required server GPUs, but WebAssembly and WebGPU are moving some of it into the browser. Model size is the constraint: a capable model can be tens or hundreds of megabytes to download. Anything genuinely collaborative — shared editing, sync across devices, sending a file to someone else — needs a server, because the whole purpose involves another machine.
What this means in practice
The practical value goes beyond privacy. Tools that run locally have no upload or download wait, which makes them substantially faster on large files — a 50 MB PDF that takes a minute to upload and a minute to download processes in seconds locally. They have no server-imposed size caps, no daily usage limits, no queues when the service is busy, and no watermarks on a free tier, because there is no marginal cost to the operator for each operation you perform. They also keep working on a train, on a plane, in a basement, and when the operator's servers go down. A local tool that you have loaded once will keep functioning regardless of what happens to the company that published it, which is a durability property server-side tools cannot offer. The trade-off is that your device does the work, so a very large job on an old phone will be slower than the same job on a server — and a page still has to be loaded once before any of this applies.
Making tools available offline deliberately
If you regularly work somewhere without a connection, you can go further than the accidental case. Most browsers let you save a page for offline use, and a tool that runs locally will generally work from that saved copy. On desktop Chrome or Edge, use the install or 'create shortcut' option where a site offers it, or simply keep the tab open — a loaded page keeps working through a connection drop. On mobile, 'Add to Home Screen' achieves something similar for sites that support it. The reliability of this varies by how the site is built, and it is worth testing before you depend on it: load the tool, disconnect, and confirm it still runs before you are somewhere without signal and actually need it. The broader point is that the offline test is not a party trick. It is the clearest available evidence about where your data goes, and it takes about fifteen seconds to perform on any tool you are considering trusting.
Tools mentioned in this guide
Frequently asked questions
How can a website work without internet?
- The network is needed to load the page. Once loaded, the JavaScript runs on your own machine — so a tool that processes files locally needs no further connection.
Does working offline prove my files are not uploaded?
- Yes. A tool that transmits your file cannot function without a connection. If it completes the operation offline, the processing happened on your device.
Which tools cannot work offline?
- Anything fetching external data — thumbnail downloaders, live currency rates, domain lookups — and anything collaborative, since those involve another machine by definition.
Are offline-capable tools slower?
- Usually faster, because there is no upload or download wait. On very large files an old device may be slower than a server would be, but the network round trip is typically the bigger cost.
Can I save a tool for offline use?
- Often yes, via your browser's install or add-to-home-screen option, or simply by keeping the tab open. Test it before relying on it, as support varies by site.
Do offline tools still get updates?
- They update when you reload the page with a connection. A cached copy keeps working as it was until then.
Last updated:
Written by Ahmed Kobbi, who builds and maintains SwiftTooly.