You open an online grocery store and drop a box of cereal into your cart. At the same moment, someone a few neighborhoods over adds the same cereal from the same store. The product page looks identical to both of you. And yet the price in the cart is different. One of you pays a little more. No one tells you so.
In December 2025, the U.S. consumer group Consumer Reports, working with two advocacy organizations, showed that this scene is really happening. More than 400 volunteers across the country simulated Instacart orders and collected prices for the same items, and about three-quarters of the products checked were shown at different prices depending on who was looking. On everyday goods like cereal, pasta, and peanut butter, the gap between the lowest and highest price ran as high as 23% in some cases. Consumer Reports gave the practice a name: surveillance pricing.
Let’s be clear about one thing. What surfaced here isn’t quite “reading your personal data to aim a price at you alone.” It’s closer to an algorithmic experiment that throws different prices at different shoppers at the same time, and Instacart pushed back on the surveillance-pricing label itself, saying it doesn’t set prices by demographics or individual behavior (and it discontinued the experimental technology in question heading into 2026). What Consumer Reports was warning about wasn’t today’s personalization but what happens when this experiment grows into a stage where it reads your data and prices each person differently. To see where that leads, we have to reach for an extreme that economics sketched a hundred years ago.
The extreme economics drew long ago
Charging different customers different prices has a name in economics: price discrimination. The word “discrimination” is there, but it’s a technical classification, not a moral charge. The early-20th-century economist Arthur Pigou sorted it into three kinds, and that framework still sits in the textbooks today.
The most familiar kind divides people into visible groups and charges each a different price. Student discounts, senior rates, regional pricing all belong here. A slightly more refined kind gets the customer to sort themselves. Coupons are the classic case. Someone willing to go through the trouble of clipping a coupon is signaling they’re price-sensitive, so you sell only to them at a discount. The seller never has to figure out who’s sensitive. The customer classifies themselves through their own behavior.
And the extreme is perfect price discrimination. Economists call it first-degree price discrimination: the seller works out exactly the most each individual customer is able to pay, that is, how much they’re willing to pay (economists call this their willingness to pay), and charges precisely that. It extracts that maximum down to the last cent.
This picture shows why perfect price discrimination is the extreme. When a single price is set, anyone who would have paid more comes out ahead. If you were prepared to pay $10 but bought it for $5, that $5 gap stays in your pocket. Add up all those gaps and you get what’s called consumer surplus. Perfect price discrimination sweeps that surplus away entirely. Charge each person their own maximum, and no gain is left over. Textbooks always attach the same caveat when they introduce it: in the real world it’s nearly impossible, because the seller has no way of knowing what’s in the customer’s head, namely how much they’re actually willing to pay.
That very “no way of knowing” is crumbling in the face of data.
This is nothing new
Charging different people different prices is itself an old practice. Before the fixed price tag, prices in the market were set by haggling. The merchant read the customer’s clothes, speech, and hesitation, and named a price. That was, crudely, personalized pricing.
More refined forms have been around a long time too. U.S. college tuition is one. The sticker price is set high, and after each family reports its finances on a financial-aid form, the actual amount each student pays is set differently. Economists see this not as soft charity but as plain price discrimination, because the aid form is the very device for learning willingness to pay. The high tuition paid by well-off families covers the lower tuition of families with less.
So personalized pricing is no invention. What AI changed isn’t that prices vary between people, but the way they’re figured out. Three things shifted. Scale shifted: where a haggling merchant read three or four cues, there are now hundreds of signals. Speed shifted: the price is recalculated with every click. And above all, the symmetry broke. In the give-and-take of the market, the merchant watched the customer and the customer watched the merchant. Now only one side watches. You don’t even know you’re standing on the scale.
What signals do they weigh you by
So what does a company actually look at to set a price? The U.S. Federal Trade Commission, a consumer-protection agency, dug into how far this practice had spread. In July 2024 it ordered eight intermediary firms that handle pricing for various retailers to disclose what data they use and how, and in January the following year it released preliminary findings. The list of signals in it hints at the markings on this scale. Precise location, search and purchase history, even the path a mouse traces across a web page, and the items left in a cart but never bought. The more they’ve learned about you, the logic goes, the more narrowly they can estimate how much you’d pay.
An old case shows the principle clearly. In 2012, it came out that the travel-booking site Orbitz was showing Mac users more expensive hotels. It was based on data that Mac users spent, on average, $20 to $30 more per night than Windows users and were more likely to book high-end hotels. To be precise, though, Orbitz didn’t sell the same room to Mac users at a higher price; it placed the pricier rooms more prominently. It changed the view, not the price. Even so, the case gets cited often because it showed how a trivial signal like the device you use can hint at the size of your wallet.
Subtler signals like battery level have long been talked about too. The idea is that someone whose battery is nearly dead is in a hurry and would swallow a premium more readily. There’s no confirmation that anyone actually prices this way, and the companies involved deny it. Even so, California introduced a bill in 2025 to flat-out ban raising prices based on battery level or device type. It judged this something that could happen, even if it hasn’t yet.
The price tag was a brief exception
Step back and an interesting picture appears. The fixed price we take for granted, the principle that goods carry a price tag and that it’s the same for everyone, isn’t actually that old.
Well into the mid-19th century, shop prices were set by haggling. In 1861 in Philadelphia, John Wanamaker opened a store under the banner of “one price, and goods returnable.” A Quaker, he held that just as all are equal before God, all should pay the same price. Wanamaker didn’t invent the price tag, but as he popularized it and hardened it into the standard of commerce, haggling was pushed into the back room.
Widen the lens and the era of price tags, where everyone pays the same, may be a mere 150-year exception in the long history of commerce. The thousands of years before it were an age of haggling. In that light, surveillance pricing is less a new monster than a drift back, past a brief pause of fixed prices, toward the original haggling. Only this time there’s one decisive difference. Only the side naming the price can see the other.
Why we can’t stand this particular unfairness
If it’s only a return to a world of haggling, why do we find this haggling so hard to bear? People react less to a price going up than to the reason it goes up. A 1986 experiment by Daniel Kahneman and two colleagues showed this cleanly. The morning after a snowstorm, a hardware store raises the price of a snow shovel from $15 to $20. How does that sit with you? Asked in a survey, 82% of respondents called it unfair. The problem wasn’t the raise itself but that it exploited someone’s urgent need. People accept a price going up because costs went up. But charging more because the other person is cornered feels like exploitation.
Surveillance pricing lands exactly on this spot. Charging you more because you look like you can pay more is precisely the kind of price hike people find hardest to stomach. And on top of that, the other side hides it.
We saw how this discomfort erupts back in 2000. Amazon was caught running an experiment that sold the same DVD to different customers at different prices. When one user cleared their browser cookies, a price that had been $26.24 dropped to $22.74. Amazon explained that it wasn’t aimed at particular customers but was a test that randomly varied the discount. Yet in the face of having paid a different price than others for the same thing, how the price was arrived at didn’t matter. The backlash was fierce, and Amazon apologized, refunding 6,896 people an average of $3.10 each. The word people reached for in their anger was closer to “deceived” than “unfair.” That the price differed hurt less than that they hadn’t known. Amazon has said, for more than twenty years since, that it doesn’t use personalized pricing.
But is it simply bad?
The anger is fair enough. Still, one thing has to be said honestly. Personalized pricing doesn’t only ever fleece the consumer.
One economics study (by Jean-Pierre Dubé and Sanjog Misra) analyzed a real field experiment and found that under personalized pricing, more than 60% of consumers actually paid less than they would have under a single price. That’s because you can name a lower price to those with a lower willingness to pay. Bringing in people who can’t afford much by lowering the price for them follows the same logic as college tuition scaling with means. So from a welfare standpoint, it’s hard to call personalized pricing bad in only one direction.
Let’s nail down the direction, though. Even if many people pay less, sellers extract more from the well-heeled few, so consumers as a whole lose out. In the same study, total consumer surplus fell about 23% compared with a single price, because the gains of those most willing to pay pass to the seller.
One common confusion is worth clearing up too. People sometimes call it surveillance pricing when airfares or ride-hail fares rise and fall with time and demand, but the two are different. Dynamic pricing that climbs in peak season is the same price for everyone at that moment. Surveillance pricing is a different price for different people at the same moment. Defenders often wrap the latter in the legitimacy of the former: isn’t it natural for prices to rise when things get busy? But things being busy and you looking pressed for time today are entirely different stories.
An analysis the Bank of England put out in 2026 goes a step further. There’s no evidence yet that personalized pricing has noticeably lifted overall inflation. But if prices shift from person to person moment to moment, people start expecting the price they’ll pay to keep rising, and as those who anticipate the rise rush to open their wallets and sellers raise prices to match, that expectation can push actual prices up on its own, the warning goes. In fact, 44% of surveyed households said they expected the price they’d pay to go up because of these corporate pricing technologies. Beyond the individual wallet, the shared sense of what a price even means begins to wobble.
The scale is already tilting
Surveillance pricing isn’t a finished future. The FTC probe lost momentum at the preliminary stage as administrations changed, Instacart shelved its experiment, and companies all say they don’t price by personal data. The full realization of the perfect price discrimination economics drew remains a theoretical extreme. What’s in a person’s head can’t, in the end, be fully read even from data.
But the direction is clear. In the U.S., by mid-2026, more than 24 states had introduced over 40 related bills. New York required disclosure when algorithmic pricing is used, and Maryland moved to ban it outright. That regulation is rushing like this means everyone can feel which way the scale is already tilting.
Maybe what really unsettles us isn’t that the price goes up. As we saw, personalized pricing hands some people a cheaper price, and the majority may come out ahead. And yet the reason we can’t stand it, as the snow-shovel experiment shows, isn’t the price itself but the feeling that our circumstances were used against us, and the fact that we alone can’t see the scale turning.
The gaze that weighs a price has always been there. Even in the haggling market, the merchant sized you up. What changed is that the gaze now runs one way. The other side reads your location, your habits, your hesitation, while you can’t even see what number is written on your own tag. What we need to reclaim, in the place the price tag left empty, may not be a cheaper price but the right to know that we’re on the scale at all. And then the question is this: can we make that scale one we can look back at too?