Browser shopping extensions have quietly become one of the largest and least-examined corners of consumer software. A handful of them sit installed on hundreds of millions of devices, watching essentially every purchase a household makes. The product they sell looks like "free coupons" or "free cashback." What they actually sell, almost always, is everything else.
This post is a sober tour of how the model really works in 2026, with the math, the data flows, and the parts the marketing pages skip over.
Those numbers are aggregated from public S-1 filings, industry research, and our own trackers of partner-program disclosures. They have shifted by maybe a percentage point over the last three years. The shape of the industry has not.
Every cashback or coupon extension sells two products at the same time. The one you see is the small commission rebate or the auto-applied coupon code. The one you don't is your shopping graph: a longitudinal record of what you looked at, what you put in carts, what you actually bought, at what price, and how often. Each of those records is far more valuable to a brand or a market-research firm than the cents it cost to send your half-euro of cashback.
Even the few extensions that don't directly sell raw history will license aggregated panels: "shoppers in this segment, in this region, between these dates." The aggregation is doing less work than it sounds. With panel sizes in the millions and purchase-event granularity at the minute, the data is re-identifiable almost trivially.
To their credit, the largest players have been forced into more disclosure over the last 18 months. Some have stopped hijacking affiliate cookies set by content sites. Some now mention, in the install flow, that they read every page you visit. None of them have changed the underlying economic model, they couldn't without restructuring the business, but the surface has politely softened.
You can usually find the truth in the EU privacy disclosures: search for phrases like "anonymised purchase metadata," "licensed to commercial partners," or "transaction-level data shared for research." Those phrases are doing very heavy lifting.
A non-broken model has, at a minimum, these properties:
Whether or not you switch shopping extensions, there are still useful things to do today.
First, audit which extensions actually have permission to read every page you visit. In Chrome that's chrome://extensions; in Firefox it's about:addons. You will probably find at least one cashback or coupon extension you forgot you installed three years ago.
Second, when you do install something new, read the privacy policy specifically for the words "share," "license," "aggregate," and "research partners." If those words don't appear, the policy is either honest or sloppy. Either way, it's a useful signal.
Third, treat the cashback rate as the upper bound of the value you'll get. The hidden cost, the slow leak of your purchase history into the brokered-data economy, almost always exceeds the explicit reward, especially at the 0.5–2% rates that are typical of the category.
None of this is meant to be moralising. The category exists because it works for both sides, in a narrow sense. But "narrow sense" is doing a lot of work.
You're all set. Click the Trackbee icon in your toolbar to start tracking.