Why the claim stopped being useful
A few years ago, saying that a photograph was analysed in the browser and never uploaded set a tool apart. Now most of the sites in this category say some version of it, and the sentences look identical whether they are describing a strict architecture or a server that deletes files after a few hours.
Those two things are not close to the same. One means the picture never existed anywhere except the machine in front of you. The other means it existed on somebody else's machine, was readable while it was there, and was removed on a schedule you are trusting them to keep. Both can be described with the same marketing sentence.
So do not read the sentence. Watch the network.
The check itself
Every desktop browser ships with a panel that lists every request a page makes. You do not need an extension, an account or any technical background. You need to be able to read a list.
- Open the tool, but do not add a picture yet.
- Press F12, or right click the page and choose Inspect. Then open the Network tab.
- Click the small clear icon to empty the list, so you are looking only at what happens next.
- Now add your photograph and let the tool do its work.
- Read the list. Sort by size if the column is available.
Safari hides this behind Settings, Advanced, Show features for web developers. Firefox and Chrome both open it with F12.
How to read what you find
If the picture never leaves, the list stays boring. You will see requests for scripts, styles, fonts and perhaps a font file or an icon, all of them small and all of them going out rather than coming back with your data. Nothing in the list will be the size of your photograph.
If the picture is uploaded, it is difficult to miss. A photograph from a phone is somewhere between two and eight megabytes, and it has to travel as one request. Look for a POST or a PUT whose size is roughly the size of your file. That is your picture leaving.
Two things commonly confuse people. A large download is not an upload, so a big response coming back is usually a script bundle and is fine. And a tool that uploads a finished result, such as a PDF you asked to be printed, is doing something different from a tool that uploads the source photograph. Look at what is being sent and when.
What you will find here
Run the check on Pacyra and the list stays boring. The photograph is decoded, cropped, analysed, segmented and exported in browser memory, and there is no address on this site that accepts one. That is not a policy decision that could be quietly reversed in a release. There is no route to send it to.
There is one exception and it is worth stating precisely, because a tool that claims no exceptions is usually the one that has not looked. If you order a printed canvas, the browser uploads two files it drew itself: the print image and the guide that goes with it. Those are needed to make a physical object and post it to you. The source photograph is no part of either file, and it is still never sent.
The act that starts that upload is adding the canvas to your cart. Before you press it nothing has been drawn and nothing has been sent.
When any of this actually matters
For a photograph of a landscape, it does not matter much, and it is reasonable to say so.
It matters when the picture is of a person who did not choose the tool. A photograph of a pupil, a client, a patient or somebody else's child is not yours to hand to a third party in exchange for a worksheet, and the question of what happened to it afterwards is one you may have to answer to somebody.
That is the case this check is for. Run it once on whatever you use, and you will know rather than trust.