Your privacy settings cost you challenges
This is the row nobody wants to read, so it is written plainly: the strictest anti-fingerprinting modes make you more distinctive to a bot check, not less. That is the trade. Here is what it actually buys and what it actually costs.
If you turned on a strict privacy mode and the captchas began, those two facts are related, and the direction is the opposite of what the settings screen implies. The protection is real against tracking; against a bot check it works against you.
Resist-fingerprinting modes work by making everyone report the same values. Time zone becomes UTC. Screen size rounds to a fixed grid. Pixel ratio pins to 1. Core count clamps to 2. Fonts collapse to a small standard set. The intent is that if everybody looks alike, nobody can be followed.
Against a tracker, that reasoning holds. Against a bot check, it inverts — because the bot check is not trying to recognise you across sites. It is asking whether this request looks like the ordinary run of requests, and the ordinary run of requests has real time zones and odd screen sizes and eleven cores.
The combination, not the values
Each of those values is individually common. Plenty of real machines report exactly two cores. Plenty of real people are in UTC. What is uncommon is all of them at once, in one request, alongside a coarsened timer and a refused storage write.
That is why your reading fires this row on two or more tells rather than on any single one, and why it prints which ones fired. A single tell is ordinary and this page says so.
Which settings fire this row, one at a time
Five tells are read, and knowing which of yours fired is the difference between one checkbox and a general anxiety about privacy settings. Your reading names them individually; here is what each one is.
- A time zone forced to UTC. Your browser reports UTC while your address, your language and your habits belong somewhere else. This is the loudest single tell here, because it also collides with the geography comparison and fires two rows at once.
- A screen size rounded to a round number. Real displays have awkward dimensions and real windows are dragged to awkward sizes. A window whose reported size lands exactly on a hundred-pixel grid, repeatedly, is a rounded value rather than a measured one.
- A pixel ratio pinned to 1. Most modern laptops and phones are not 1. A reported 1 on a device whose other properties suggest a high-density display is a contradiction rather than an oddity.
- A core count clamped to 2. Common on genuinely small machines, and it is only weighed here when it arrives next to the others.
- A coarsened timer. Some browsers deliberately reduce how precisely a page can measure elapsed time. It is invisible to you and unmistakable to anything that times itself, and it is corroboration rather than a cause in its own right.
The narrower setting
In Firefox, the practical answer for most people is Enhanced Tracking Protection on Standard or Strict, with privacy.resistFingerprinting left off. You keep tracker and cookie blocking — which is what actually stops the following — and you stop reporting a machine that does not exist.
In Safari, Intelligent Tracking Prevention is on by default and is not the problem. Advanced Fingerprinting Protection is a smaller change than Firefox's and rarely fires this row on its own.
In the Tor Browser, this row will always fire and should. Everything above is deliberate there and the challenges are the known price.
Is Tor Browser always going to be challenged?
Yes, and it is worth understanding that the answer has two independent halves so you do not spend an afternoon on the wrong one.
The first half is this row: everything Tor Browser does to make its users identical to one another is, to a check asking whether a request looks ordinary, a set of answers almost nobody else gives. That is the design working as intended and there is no setting that keeps the protection while removing the effect.
The second half is the exit node, which is a shared address on a published network carrying a very large amount of other people's traffic. Two of the heaviest rows on this site fire simultaneously and neither has a fix inside the browser.
The honest position: if you use Tor for reasons that matter, the challenges are the cost, and no page including this one can remove them. If you are using it casually and finding the web unusable, an ordinary browser with tracker blocking gets you most of the practical privacy at none of this price.
Do private windows help or hurt?
Neither, mostly, and that surprises people who reach for a private window when a site starts refusing them.
A private window does not change what your browser reports about the machine. Same time zone, same screen, same fonts, same renderer, same address. What it changes is storage: cookies and local data are discarded when the window closes, and third-party storage is frequently restricted more aggressively inside it.
For a bot check that gates on a stored pass, that is strictly worse — you cannot keep a pass you discard, which is exactly why a captcha loops. Opening a private window is one of the most common reflexes when a site starts challenging you, and one of the few that reliably makes the specific symptom worse.
Randomisers are worse than either
Extensions that randomise the canvas, the user agent or the screen size on every page load produce something no real machine produces: a browser whose own answers change between one request and the next. A vendor does not need a population to notice that. It needs two requests.
If you have one installed, turning it off is usually the single largest improvement available to you, and it is the one this page will recommend before any other hardening change.
What actually protects you, at no cost here
It would be a poor page that told you privacy tooling costs you challenges and stopped there, so: the measures that stop cross-site following are mostly not the measures that fire this row.
Blocking third-party cookies, blocking known tracker hosts, isolating storage per site, and not being signed into a large advertising network while you browse all reduce what can be assembled about you across sites. None of them change what your browser reports about the machine, so none of them appear on this row. The settings that do appear are the ones that lie about the machine — and lying about the machine is a defence against a fingerprint, not against a tracker.
That is the whole trade, and it is a narrower one than the settings screens make it look.
Privacy hardening
Your anti-tracking settings make you stand out instead of blending in.
- Move from the strictest anti-fingerprinting mode to the standard one. You keep tracker blocking; you stop reporting a machine nobody else has.
- If you use resist-fingerprinting deliberately, leave it on and expect challenges. That is the trade, stated plainly.
- Turn off any extension that randomises canvas, screen size or user agent — a value that changes on every reload is worse than a value that is merely unusual.
A row's place on your reading is one multiplication, and both halves of it are published. Neither half is a measurement.
- Surprisal −log₂ of an assumed probability of 0.0300 for “a resist-fingerprinting tell is present”
- 5.06 bits
- Challenge weight Strict anti-fingerprinting settings — how much a bot check is judged to act on it
- 0.50
- Rank The product. It orders the rows and nothing else.
- 2.53
Why that assumption
A time zone forced to UTC, a rounded screen, a pixel ratio pinned to 1 and a core count clamped to 2 are individually plausible and jointly rare. The prior is for any one of them appearing.
Why that weight
Resist-fingerprinting modes pin the time zone to UTC, round the screen, clamp the core count and flatten the pixel ratio. The values are common; the *combination* is not, and it is a combination almost nobody arrives at by accident.
Every assumption, with its number → · Every weight, and what is deliberately not measured →
Questions about the “Privacy hardening” row
- Why do I get more captchas with privacy settings turned on?
- Because a bot check is not trying to recognise you — it is asking whether your request resembles the ordinary run of requests. Hardening makes your browser report a set of values chosen for uniformity: a UTC clock, a rounded screen, two cores, a coarse timer. Individually those are ordinary. Arriving together in one request, they are a combination almost nobody reaches by accident.
- Should I turn off resistFingerprinting in Firefox?
- For most people, yes, and keep Enhanced Tracking Protection on. Tracker and cookie blocking is what actually stops you being followed between sites; resistFingerprinting is a defence against a different threat and it is the part that makes your browser describe a machine that does not exist. Turning it off costs you very little of the protection and removes most of this row.
- Does using a private or incognito window reduce bot checks?
- No, and it commonly makes one symptom worse. A private window reports the same time zone, screen, fonts and address as an ordinary one, so nothing a check reads about the machine changes. What does change is that stored data is discarded — including the pass a challenge issues you — which is the direct cause of a captcha that keeps returning.
- Will Tor Browser always be challenged?
- In practice yes, and for two separate reasons that stack. Everything it does to make its users identical is a set of answers a check reads as unusual, and its exit nodes are heavily shared addresses on networks published as infrastructure. Neither has a fix inside the browser. If you need Tor, that is the cost; if you were using it casually, ordinary tracker blocking gets you most of the day-to-day privacy without it.
- Are anti-fingerprinting extensions that randomise values a good idea?
- They are the worst option on this page. Randomising the canvas, user agent or screen size means your browser gives different answers to the same question on consecutive requests, and no real machine does that. An unusual value is merely unusual; answers that change between one page load and the next are evidence that something is rewriting them, and that is acted on much harder.
READ AT PRIOR TABLE ED. 1.0.0 · 2026-10-13 · ASSUMED, NOT MEASURED
The rank is taken against 15 written assumptions about how common each value is and 16 written weights about how much each one matters. 9 of them we think are probably wrong. Real counts from the browsers these sites have read appear beside a row, never inside its rank. Read the assumptions →
The rows you cannot fix from a settings page
A VPN you need for work, an address your ISP shares with a thousand other subscribers, a laptop with no graphics chip, a hardened profile that breaks the sites it was meant to protect — none of these have a checkbox. What is left is to present a reading that is internally consistent and ordinary on purpose, instead of one that is accidentally strange.
Donut Browser does that at the engine level: its Wayfern Chromium fork applies the fingerprint through CDP injection rather than patching JavaScript in the page, and it refuses to launch a profile whose proxy exit country, time zone and language contradict the profile's own fingerprint — which is the top row of most readings on this site. It is AGPL-3.0, it has no telemetry of any kind, and it runs on macOS, Windows and Linux.
Stated plainly: this site has no test results showing what any particular anti-bot service does with any particular browser, and neither does Donut Browser's own repository. What can be checked is the source, which is public.
Free · AGPL-3.0 · github.com/zhom/donutbrowser