The Rise and Fall of Google's Privacy Sandbox
This paper presents the first longitudinal, consent-aware measurement of Google's Privacy Sandbox, revealing that its APIs suffered from stagnant adoption and limited ecosystem support well before their announced retirement in October 2025, ultimately leaving the challenge of privacy-preserving interest-based advertising unresolved.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the internet as a massive, bustling city where every shop owner wants to know exactly what you like so they can show you the perfect sign for your next purchase. For years, the way they did this was by following you from store to store with a tiny, invisible sticker on your backpack—this is what tech people call a "third-party cookie." It was effective, but it made many people feel like they were being watched, leading to a big push for privacy. In response, Google, the giant who runs the city's most popular map, tried to build a new system called the "Privacy Sandbox." Think of this as a fancy, high-tech security guard who promises to let shops know your interests without ever letting them see your face or follow you around the city. The idea was to move the tracking work inside your own browser (your personal phone or computer) so it stays private, but the system was so complicated that it needed to be redesigned many times. The big question everyone was asking was: Did this new security guard actually work, or did the shops just ignore it and keep using their old, creepy stickers?
This paper is like a team of detectives who spent six months secretly watching the top 10,000 websites in the city to see how this new system was actually being used. They didn't just look at the numbers; they watched what happened when you said "yes" to privacy, when you said "no," and even when you hadn't clicked anything yet. They found out that the Privacy Sandbox was a bit of a flop. Even before Google officially announced they were shutting most of it down, the shops had already stopped using the new security guard. The only part that stuck was a specific tool called CHIPS (Cookies Having Independent Partitioned State), which is like giving each shop a separate, locked box for your data so they can't compare notes across the city. But even with this one success, the old, invasive stickers are still the most popular way to track people. The study shows that the ecosystem of advertisers and websites simply wasn't ready to switch to this new, privacy-friendly way of doing things, leaving the problem of how to show you relevant ads without spying on you still unsolved.
The Rise and Fall of the Privacy Sandbox
So, what exactly did Google try to build, and why did it crumble? The Privacy Sandbox was supposed to be a magic box that let advertisers show you ads based on your interests without needing to know your name or track your every move across the web. Instead of a single, long-lasting sticker (a third-party cookie) that follows you everywhere, Google proposed a suite of new tools. Some tools would let your browser guess your interests locally (like the Topics API, which gives you a generic label like "sports fan" instead of a list of every sports site you visited). Others would let you see ads for things you looked at before, but only by running a secret auction inside your browser (the Protected Audience API).
The researchers, a team from Italy, decided to test if this magic box was actually being used in the real world. They built a special robot crawler that visited the top 10,000 websites every week for six months. This robot was smart: it could pretend to be a user who hadn't clicked "Accept" on a privacy banner yet, a user who said "Yes," and a user who said "No." This allowed them to see if websites were playing by the rules or sneaking in tracking before you even agreed to it.
The Big Discovery: A Quiet Abandonment
The most surprising thing the team found was that the Privacy Sandbox didn't just fail; it was quietly abandoned long before Google officially pulled the plug. In October 2025, Google announced they were retiring most of these new tools. But the data showed that the usage of these tools had already been shrinking for months.
Think of it like a new, eco-friendly car that a big company promised would replace gas cars. The company said, "We'll stop making gas cars next year!" But the data showed that people had already stopped buying the eco-car six months ago because it was too complicated and the gas stations (the old tracking methods) were still working fine.
The study found that:
- Most tools were ignored: Out of ten different tools in the Privacy Sandbox, only one—CHIPS—was used by a significant number of websites (about 50%). The others, like the Protected Audience and Shared Storage, were used by almost no one. In fact, their usage dropped to nearly zero months before the official retirement announcement.
- Only a few tried it: The few websites that did use these tools were mostly giant advertising companies. But even they started dropping the tools as time went on. By the end of the study, the number of companies using these tools had shrunk significantly.
- The old ways won: Despite all the effort to build a privacy-friendly system, the old-fashioned, unpartitioned third-party cookies (the "creepy stickers") were still being used by about 75% of websites. Even the one tool that worked (CHIPS) didn't replace the old cookies; it just sat alongside them.
The "Questionable" Behavior
The researchers also looked at what happened when users said "No" to tracking. They found that some websites were still using these new tools even after a user explicitly denied consent. It wasn't necessarily a malicious conspiracy, but rather a sign that the system was messy. It's like a store that promises to stop following you if you say "stop," but sometimes their security guard forgets to turn off the camera. This "questionable usage" showed that the rules for privacy were hard to enforce automatically.
Why Did It Fail?
The paper suggests that the Privacy Sandbox failed because it was too complex and didn't offer a clear benefit to the people running the websites. The old way of tracking (third-party cookies) was simple and worked well, even if it wasn't great for privacy. The new system required websites to change how they built their pages, and the tools were often confusing or didn't give advertisers the data they wanted.
The study concludes that just having a "good idea" or a big company like Google pushing for it isn't enough to change how the internet works. The ecosystem of advertisers and websites is too stuck in its old habits. The researchers suggest that maybe only strict laws (regulations) could force a real change, but for now, the dream of a privacy-preserving internet where you can still see relevant ads remains unfulfilled. The Privacy Sandbox was a noble experiment, but in the end, the internet decided to keep using the old, invasive methods.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.