The Price Tag is Watching You
Walmart is installing digital price tags in every U.S. store. Some retailers are already scanning your face at the door. The infrastructure for connecting those two systems is being built faster than
Electronic shelf labels and in-store facial recognition are spreading through the country’s largest grocery chains on separate tracks right now. Regulators are only just starting to look at what happens when those tracks meet.
“Two people standing in the same store, buying the exact same item, at the exact same time, could be paying two different prices based on what they’ve told their chatbot. Why? Because an algorithm decided one of them could afford more.” — Congressman Josh Gottheimer (NJ-5), May 18, 2026
Three Things to Know
Walmart will have digital price tags, changeable in real time, in every U.S. store by the end of 2026, and the company holds two 2026 patents for AI systems that automatically update and recommend prices.
When the ACLU asked America’s 20 largest retailers whether they scan customers’ faces, 17 refused to answer. Wegmans confirmed it does, after a legally required notice went viral.
Maryland and Connecticut have now banned surveillance pricing in groceries, but both exempt loyalty program pricing, the exact mechanism that builds the profile on which a personalized price would run.
The signs went up quietly. Small notices, required by New York City law, appeared near the entrances of Wegmans stores in Manhattan and Brooklyn in January. They stated, in clinical terms, that the company collects, retains, converts, stores, or shares customers’ biometric identifiers, including facial recognition, eye scans, and voiceprints. Customers who’d shopped there for years had no idea. The story went viral within days.
Wegmans confirmed the signs were real. The company said it uses facial recognition in a small fraction of stores in communities at “elevated risk” for security to identify people already flagged for misconduct. That part may be true. What the story actually revealed wasn’t Wegmans. It was the silence around it. New York City is one of the only places in the country that requires a business to tell you when it’s scanning your face. So the ACLU asked 20 of America’s largest retailers directly: Do you use facial recognition on customers? Seventeen refused to answer. Lowe’s confirmed it does. Walmart, Kroger, and Home Depot acknowledge the technology buried in their privacy policies, which almost nobody reads.
The Infrastructure Being Built
Electronic shelf labels aren’t new. What’s new is the scale and the speed. Walmart will have them in every U.S. store by the end of 2026, up from roughly 2,300 stores earlier this year. Kroger, Whole Foods, Amazon Fresh, Schnucks, and Lidl have already deployed them chainwide. The stated rationale is efficiency, and it’s real: a price change that used to take an employee two days to complete across a 120,000-item store now takes minutes via a mobile app.
Here’s where I stop taking the efficiency framing at face value. In January 2026, Walmart received a U.S. patent for a system that “dynamically and automatically updates item prices.” In March, it received a second patent for using machine learning to forecast demand and recommend prices at scale. Walmart says publicly that it will never use this for surge pricing, and a spokeswoman has said the price you see is the same for every shopper in a given store. I believe that’s true today. What I’m pointing at is the gap between that promise and the patent filings sitting next to it. A company doesn’t usually patent a capability it has no intention of ever using.
The United Food and Commercial Workers union, which represents 1.2 million grocery workers, launched a national campaign in February specifically to stop this technology before it gets used for what its critics say it was always built to enable.
The Circuit
Electronic shelf labels are one piece. Facial recognition cameras are a second. Loyalty programs are the third, and they’re the one most people don’t think to be suspicious of, because loyalty cards have existed for decades as a discount mechanism. Structurally, that’s not what they are. A loyalty program is a data collection system offering a discount as its user interface. Every scan builds a profile: what you buy, what you almost buy, when you shop, how price-sensitive you are for specific items, and which promotions actually move you. That profile has commercial value on its own. It becomes worth more the moment it can be attached to a face.
Kroger announced plans to use facial recognition for targeted coupons, personalizing prices based on who you are. Customers pushed back hard enough that Kroger said it had no current plans to identify faces at digital displays. Notice what that statement does and doesn’t say. It addresses one specific application. It leaves the rest of the infrastructure exactly where it was.
Here’s what the full circuit looks like once it’s assembled, and I want to be precise that this is the architecture being built, not a system that fully exists in any one store today: a loyalty app identifies you the moment you walk in. Facial recognition cameras confirm it and log what you stop in front of, how long you hesitate, what you pick up, and what you put back. The electronic shelf label in front of you shows a price calibrated to that profile. The person standing next to you, buying the same item at the same moment, may see something different.
Running This Through the Four-Part Test
I have a framework I use across this site to test whether something is actually coercive, not just unfair or unpleasant. I call the broader pattern Coercive Capitalism: systems that don’t force you into anything outright, but engineer your compliance by exploiting the gap between who you are and who you’ve been told you should be. The test itself has four parts, and something has to clear all four before I’ll call it coercive rather than just a bad trade-off. Is the trade voluntary? Did you receive a genuine benefit in return? Is the data you handed over weaponized against you, resulting in a real financial or material cost? And is the cost of walking away engineered to be prohibitive, rather than just inconvenient?
Run a loyalty card through it. Signing up is voluntary, and the discount is a genuine benefit, so it clears the first two parts cleanly. The third part is where it turns. The same data that earns you 40 cents off cereal is the raw material for a system that can, eventually, decide you’ll tolerate paying more for it. That’s the data being weaponized back against the person who generated it, just routed through a different mechanism than the ones I usually write about here. The fourth part, exit cost, is what makes grocery shopping a different category entirely from airline tickets or hotel rooms. Nobody needs a hotel room. Everybody needs to eat. “Just shop somewhere else” stops being a real option when the largest cha
Who Actually Pays More
This isn’t evenly distributed. The data inputs that feed a personalized pricing model, proximity to competitor stores, purchase history, inferred income, and geography aren’t neutral variables. They map onto the same patterns that have structured economic disadvantage for generations. A system built to charge what it predicts you’ll tolerate will charge people in lower-income zip codes more for the groceries they can least afford to pay for.
Rite Aid is the warning shot for what can go wrong when this kind of system runs unchecked. The FTC banned Rite Aid from using facial recognition for five years in 2023 after finding its system generated false identifications that disproportionately flagged Black and Asian customers as shoplifters, deployed mostly in stores located in communities of color.
What the Industry Gets Right
I want to give an honest account of the other side, because retailers aren’t wrong about everything. Electronic shelf labels solve real operational problems: price accuracy, inventory management, and fewer checkout discrepancies. A UC San Diego study examining more than 180 million product-level price observations across 114 stores found no evidence that electronic shelf labels have led to surge pricing in U.S. grocery retail so far.
I believe that finding. I also think it’s a backward-looking data point about a system being built in the present tense, the same distinction I’d draw about any capability that hasn’t been switched on yet. The retailers argue that personalized pricing would damage customer trust, and that argument only holds if customers know it’s happening. Right now, this industry’s default is non-disclosure, and the Wegmans story is the clearest proof of how wide that gap actually is.
There’s already a documented version of this happening at scale, and it didn’t even need facial recognition. A Groundwork Collaborative and Consumer Reports study found Instacart was showing shoppers as many as five different prices for the exact same item, at the exact same store, at the same moment, with some shoppers paying up to 23 percent more. New York Attorney General Letitia James sent Instacart a formal demand for answers in January. Instacart said it would stop running those specific item-level price tests. Its retail partners can still run their own.
Where the Law Actually Stands
This moved faster than I expected, even a few weeks ago, so here’s the current state, not the one from when this story first got reported.
Maryland was first. Governor Wes Moore signed the Protection From Predatory Pricing Act on April 28, 2026, banning surveillance pricing in food retail, effective October 1. Connecticut followed. Governor Ned Lamont signed HB 5563 on May 27, 2026, making it the second state, also effective October 1. Both laws carve out loyalty program pricing entirely, as long as enrollment is voluntary, which is precisely the mechanism the rest of each law seeks to restrict.
Colorado passed a broader, all-industry ban on May 8. Governor Jared Polis vetoed it on June 2, calling it “overly broad.” A Democratic governor in a blue state vetoing his own legislature’s surveillance pricing bill is worth sitting with. Even people sympathetic to the underlying concern struggle to draw a clear line between surveillance pricing and ordinary personalization. California’s AB 2564 passed the full State Assembly on May 27 and is now in the Senate.
At the federal level, Senators Ben Ray Luján and Jeff Merkley’s Stop Price Gouging in Grocery Stores Act would ban the use of surveillance pricing and electronic shelf labels outright in large grocery stores. Representative Gottheimer’s No Rigged Grocery Prices Act, introduced May 19, 2026, as H.R. 8895 with Republican co-lead Mike Lawler, takes a narrower approach, banning the personal-data targeting while explicitly preserving discounts and loyalty pricing. Neither bill has cleared the committee.
The honest summary isn’t “two states banned it.” It’s the two states that acted fastest that both banned the output while leaving the input untouched. Loyalty program data is what would feed a personalized price in the first place; eye scans and price tags are just the delivery mechanism. A law that prohibits the delivery mechanism while exempting the data pipeline that supplies it isn’t a weaker version of a ban. It’s a different thing wearing a ban’s name, and it passed specifically because that exemption was the price of getting industry to stop fighting it. Connecticut moving second, and moving with the same hole in the wall as Maryland, isn’t a sign the approach is converging on something solid. It’s a sign the hole is structural, not a drafting accident; anyone’s likely to close next time, either.
What I’m Watching Next
I don’t think there’s a room where a retail executive and a software vendor sat down to plan a face-scanning, price-personalizing checkout system together. I think a loyalty program made sense on its own. A facial recognition camera for shoplifting made sense on its own. A digital price tag that saves labor hours made sense on its own. Nobody had to conspire for three separately reasonable decisions to add up to an architecture capable of charging two people different prices for the same loaf of bread. That’s the part worth sitting with longer than the outrage over any single piece of it.
What I’m watching for now, and this is the one that actually matters more than any of the others: whether a single state, anywhere, passes a surveillance pricing ban that closes the loyalty-program exemption rather than writing it in. Every bill that’s made it into law so far has needed that exemption to get the industry to stand down. If that pattern holds in California, in the federal bills, in whatever comes after this, then the lesson isn’t that the law is catching up. It’s that the law has found the one mechanism it’s structurally unwilling to touch, the same mechanism that makes the rest of the architecture work in the first place. Secondary to that: whether Connecticut’s price-setting-device disclosure actually shows up on receipts the way it’s supposed to, and whether anyone tests Maryland’s carve-out in court before October 1.
What You Can Do
Look at your grocery store’s shelf tags. If paper has been replaced by small digital screens, the infrastructure for personalized pricing is already in place, whether or not it’s being used that way yet.
Check for biometric disclosure signs at your store’s entrance if you’re in New York City. If you shop elsewhere, the absence of a sign tells you nothing, since most places don’t require one.
Know what your loyalty card actually is. The discount is real. So is the profile it’s building. Decide knowingly, not by default.
If you’re in Maryland or Connecticut, watch what happens to loyalty pricing specifically after October 1. That carve-out is the loophole worth tracking.
Contact your representatives about the Stop Price Gouging in Grocery Stores Act and the No Rigged Grocery Prices Act. Both are sitting in committee right now, which is exactly when constituent pressure moves a bill.
A very thought-provoking article. I appreciate that you distinguish between what is happening today and what the technology could enable in the future. Too often discussions like this become overly speculative, but you clearly separate current evidence from potential risk. Whether or not personalized pricing becomes widespread, it's an important conversation about how emerging technologies, consumer data, and transparency should evolve together.