Turning on every privacy toggle a browser offers rarely stops the tracking it is supposed to stop. It just moves the tracking somewhere the toggle doesn’t reach. That gap — between what a person does to opt out and what actually happens to their data — is the more honest frame for any headline that promises browser-based tracking control.
An August 2026 PCMag review of Firefox’s new AI features is one of the few current data points on what a privacy-first browser posture looks like in practice. Mozilla’s own Smart Window privacy notice states that it never retains chat logs without a user’s permission, and that the data behind its on-device “Memories” feature is processed on a Mozilla server but not kept there afterward — it lives locally on the device instead. Where possible, features like tab grouping run on local models that never contact a server. Users can opt in to AI features rather than fight to opt out of them.
Mozilla’s promise on data sales is narrower than it used to be, though. After a 2025 rewrite of its terms of use drew backlash from users, the company shifted from a flat “we don’t sell your data” to a hedged version: it doesn’t sell data “in the way that most people think about ‘selling data.'” That’s a smaller promise than the one Firefox used to make, even as its Smart Window design stays genuinely opt-in.
According to PCMag, Ajit Varma, head of Firefox, said Mozilla’s approach focuses on making AI features easy to access for those interested while remaining simple to disable for those who prefer not to use them. That sentence is doing a lot of work. It concedes something the larger platforms rarely say out loud: that the default posture of most modern browsers is opt-out, not opt-in, and that reversing it is a design decision, not a technical limitation.
Compare that with the experience users have across platforms. To remove the Gemini button from Gmail, a user has to look up a guide. The friction is the point. The friction is also the business model.
The same fragmentation shows up on shared devices and family accounts. A search made under one login can trigger ads under a different one entirely. No single toggle resolves that, because browser-level controls, account-level controls, device-level controls, and app-level controls all govern different slices of the same behavior, and switching off one does not touch the others.
This is the structural problem that any headline about a browser “allowing” opt-out obscures. A browser is one layer. The advertising identifier on the phone is another. The logged-in account is a third. The data broker who bought the profile last quarter is a fourth. Opting out of tracking at the browser level, without touching the other three, is closer to closing one window in a house with the doors open.
The pattern is not new. Privacy-adjacent regulation elsewhere has followed a similar shape, shifting the compliance burden onto ordinary users — requiring adults to verify their identity, for instance, to enforce rules aimed at protecting children online. The mechanism differs. The pattern rhymes. Someone has to prove something, click something, or accept something, and that someone is almost never the party doing the collecting.
Those who have watched consent flows evolve for years observe that the shift from Do Not Track headers in the early 2010s to today’s cookie banners to tomorrow’s Global Privacy Control signals has produced more interfaces, more clicks, and roughly the same amount of tracking. The technology of consent has advanced. The economics of surveillance has advanced faster.
What Firefox’s approach hints at, and what the larger browsers have not matched, is a different question: not whether the user can opt out, but whether the product was built to collect in the first place. Mozilla’s decision to process only limited technical information, keep memories on the local machine, and support an open model ecosystem — including local models — reframes the opt-out from a setting into a design principle. Users can even connect their own models, including ones running locally.
That is a meaningfully different posture from a company announcing that browsers may now request not to be followed.
The distinction matters because language around tracking has grown loose. Announcements about “allowing” opt-out often describe honoring a signal the user must first know exists, find in a menu, activate, and hope is respected downstream. Enforcement is thin. The Global Privacy Control signal, for instance, carries legal weight in California under state privacy law but is treated as advisory in most other jurisdictions. A browser can send the signal. Whether the site on the other end acts on it depends on where the user lives, what the site’s lawyers have decided, and whether any regulator is watching.
Users in California have one of the strongest legal footings in the country for this reason. Most users likely do not know that.
The reader takeaway is smaller and more useful than any policy headline. A browser opt-out is a floor, not a ceiling. It handles one layer of a multi-layer problem. Pairing it with account-level ad settings, a device-level advertising identifier reset, and a periodic review of connected apps closes more doors than any single toggle. None of that is glamorous. None of it is a solution. It is maintenance, the way locking a car is maintenance.
The broader pattern is worth naming plainly. The privacy debate has spent a decade being framed as a question of user choice. In practice, the design of the choice — where it lives, how many steps it takes, what it actually controls — is where the outcome is decided. Firefox’s version of that choice looks different because the underlying product was built differently, even if its promises on data sales have gotten narrower. That is the part no announcement can retrofit.
Tracking opt-outs, in their current form, ask users to do the work of a system that was designed to make the work hard.
Users who turn on every privacy toggle they can find often see results for about a week. Then the ads come back, wearing a slightly different outfit.