AI grocery pricing: a digital shelf label isn't a price set just for you

An electronic shelf label lets a store replace the displayed price without sending someone down the aisle with paper tags. That can make frequent changes easier. It does not, by itself, establish that the store varies prices by demand or by who is shopping.

Dynamic pricing means changing a price as market conditions change, such as supply and demand. Personalized pricing means varying prices or offers using information about a shopper or a group. Algorithmic pricing describes software helping recommend or set the price; it can support either practice. AI can be part of that software, but neither changing prices nor pricing software began with chatbots.

There is also a simpler possibility: the screen is showing the store's ordinary scheduled price update. AP's earlier reporting on a study of one unnamed grocery chain found virtually no surge pricing before or after electronic labels arrived. That is evidence about that chain, not a guarantee about every supermarket or its future plans.

AP points to last year's Instacart investigation as a useful distinction. Shoppers saw different prices for identical products at the same stores. Instacart subsequently ended what it described as tests that randomly assigned shoppers to groups, rather than pricing from their personal information.

A shopper can object to being placed in a price experiment without claiming the experiment knew their income. The observed difference and the explanation for it need to stay separate.

The same care applies to patents. CBS reports on Groundwork Collaborative's analysis of Walmart patent applications and the group's concerns about individualized prices. A patent application is not proof that a retailer deployed the described system. Walmart denies using personal information to set individualized prices. Reporting the concern without that distinction would leave readers with a stronger accusation than the evidence supports.

In a September 25 letter, Walmart CEO John Furner says the company does not set different prices based on who customers are or the time of day, and will not do so. He says income, shopping history, urgency and estimated willingness to pay will not change the price.

The letter extends that commitment to Sparky, Walmart's AI shopping assistant: information shared with it will not be used to raise a customer's price or hide lower-priced options that meet their needs.

That is a company's public commitment, not an independent audit. It is still more specific than a promise to use AI responsibly. It names conduct customers can challenge, including quietly narrowing the cheaper choices an assistant shows them.

The Federal Trade Commission's August announcement sought comment on a proposed enforcement policy for personalized pricing. It says businesses that fail to explain how personal data is used to set a price may violate existing consumer-protection law. The agency also says it lacks authority to ban personalized pricing in all circumstances.

The proposed statement calls for clear disclosure of the fact that a price is personalized, the basis for that personalization and the types of data used. Calling this an already enacted nationwide ban would be wrong. Calling every price difference unlawful would be wrong, too.

For a shopper, the useful question is plain: did information about me affect this price, and what information? A label saying a deal is special does not answer it.

If a price looks wrong, save the listing or photograph the shelf label and keep the receipt. Note the store, time, product size and any loyalty or coupon condition. Ask the retailer to explain the discrepancy. A mismatched checkout total is a concrete problem even when its cause has nothing to do with AI.

When comparing two offers, match the item and the conditions before drawing conclusions: the same pack size, store or delivery location, membership status, coupon and collection method. A difference is worth asking about; a screenshot pair alone may not reveal why it exists.

If you use an AI shopping assistant, ask it to preserve those conditions and compare the final basket with fees included. That is a sensible task to request, not a claim that every assistant can do it reliably. Check the actual checkout. Do not assume private browsing guarantees a lower price or removes every possible pricing input.

There are useful jobs for this technology. Updating a shelf price without hours of paper replacement is one. Helping someone compare a real basket could be another. Neither requires treating a family's urgency as a reason to charge more.

I would rather see a retailer explain its pricing boundary in a sentence than send customers hunting through patents. Until that explanation is clear, be precise about the complaint. An unexpected bill deserves an answer. The presence of a screen does not supply one.