Browser fingerprint test
Sites identify you by far more than cookies. Below is the fingerprint your browser is handing out right now: canvas, WebGL, fonts, screen, timezone, WebRTC, and, more importantly, whether those signals agree with each other.
Everything here runs in your browser. Nothing is uploaded, stored, or sent to us: view source if you'd rather check than take our word for it.
Reading your browser…
This takes a second.
Why consistency matters more than hiding
Blocking or randomising one signal usually makes things worse. A browser claiming Windows while its WebGL renderer reports Apple hardware, or reporting a Berlin timezone with a US-English locale and a New York IP, is more identifiable than one that simply answers honestly, because almost nobody looks like that. What keeps two accounts apart is each one looking like an ordinary, internally consistent device.
Parallel gives every profile its own coherent fingerprint, cookies, storage and proxy, and checks the combination for exactly the contradictions listed above before the browser starts.
Download for Windows, 3 profiles freeWhat each of these signals is
The readings above are grouped the way a site that profiles you would group them. Here is what each one is measuring, and why it is worth measuring.
Canvas and audio fingerprints
Both work the same way. The page asks your browser to render something, a line of text onto a canvas or a short tone through the audio stack, and then hashes the result. Nothing is shown and nothing is played. The bytes that come back depend on your GPU, your drivers, your font rasteriser and your audio hardware, so the same instruction produces a subtly different result on a different machine. The hash is stable between visits and survives clearing cookies, which is the entire point of it.
GPU vendor, renderer and the WebGL fingerprint
The vendor and renderer strings are read straight out of WebGL, and they are usually the most specific line on this page. Something like "ANGLE (NVIDIA, GeForce RTX 4070)" narrows you to a small population on its own. The WebGL fingerprint goes further and hashes how your hardware actually rasterises a test scene. The contradiction to watch for is a renderer that disagrees with the platform: a browser claiming macOS next to an NVIDIA renderer describes a machine Apple stopped selling years ago.
Fonts
A page cannot ask for your font list, so it measures instead. It renders a string in a candidate font, renders it again in a fallback, and compares the width. Identical widths mean the candidate was missing. What you have installed follows from your operating system, your language, and the software you use, so an office suite or a design tool leaves a signature. Which fonts matters far more than how many: a machine carrying nothing beyond the OS defaults is as unusual as one carrying hundreds.
Screen, viewport and touch points
Resolution is the panel. Available area is the panel minus the taskbar or dock, which leaks both your OS and roughly how you have arranged it. Device pixel ratio separates a scaled display from an unscaled one. Touch points tells a phone from a desktop more reliably than the user agent does, because it is harder to change. A viewport larger than the screen reporting it, or a desktop resolution alongside five touch points, is the kind of mismatch that marks a browser as modified rather than unusual.
Timezone, UTC offset, locale and languages
The first two come from your system clock, the second two from your OS settings. None of them is your IP address, and that is exactly the trap: a proxy moves your IP and changes nothing else here. A browser arriving from an Amsterdam address, reporting a New York timezone, and listing en-US as its only language has told three different stories about where it is, and a site only has to compare them.
WebRTC addresses
WebRTC collects candidate addresses so that two peers can find each other directly, and for years that included addresses on your local network that the page was never given. That is how proxy users were caught. Chrome now hides local addresses behind a randomised .local hostname, so one of those appearing here is normal and is not a leak. The finding that still matters is a public address that differs from the one your traffic appears to come from.
User agent, client hints and navigator.platform
The user agent is the oldest signal here and the least trusted, because anyone can set it to anything. Client hints are the structured replacement: brands, platform and mobile, sent as separate fields. Because they reach the site from the same browser by a different route, they work as a consistency check. A user agent and a client hints platform that disagree is a stronger signal than either on its own, and what it says is not that the machine is unusual but that the string was edited.
CPU cores, device memory and device pixel ratio
These are deliberately coarse. Device memory in particular is reported in a handful of buckets rather than exactly, so that it cannot identify anyone by itself. They are not meant to. They narrow, and they combine: a core count adds almost nothing on its own and a great deal next to a GPU string.
Cookies and Do Not Track
Whether cookies are enabled is self-explanatory. Do Not Track deserves a sentence, because it backfires: almost nobody turns it on, so turning it on makes you easier to pick out, not harder. It is a request that sites are free to ignore, and nearly all of them do.
Common questions
Does a VPN or proxy change my fingerprint?
No. It changes the address your traffic arrives from. Every reading on this page is taken inside the browser and is completely unaffected by it. That is why a proxy on its own does not keep two accounts apart, and why changing proxy while keeping everything else identical can make two accounts look more related rather than less.
Does private browsing help?
It clears cookies, history and storage when you close the window, which is useful. It does not change a single signal above. The same canvas hash, the same fonts, the same GPU, the same timezone. A site that recognised you by fingerprint before will recognise you in a private window.
What about a separate browser profile?
A Chrome profile separates cookies, logins and history, and for keeping your own accounts tidy that is usually enough. It does not separate anything on this page, because the profiles share one browser, one operating system and one set of hardware. Everything measured here is identical across all of them.
Should I block or randomise these signals?
Usually not, and the section above this one explains why: the contradictions are more identifying than the values. An extension that randomises your canvas hash on every page load makes you the only visitor whose hardware appears to change between one page and the next.
Is a unique fingerprint bad?
It depends entirely on what you are doing. For ordinary browsing it mostly means better ad targeting. It becomes a problem when you need two accounts not to be connected to each other, because then a fingerprint is the thing connecting them, and it is the one piece of evidence that no amount of logging out will clear.