A federal antitrust lawsuit filed in early October alleges that McDonald uses an AI-powered pricing engine to set menu prices across United States locations, an allegation the company denies while acknowledging it gives franchisees pricing tools and recommendations. The case lands as electronic shelf labels spread through Kroger, Walmart and Whole Foods, and as grocers adopt AI to mark down perishables. Bank of England economists warned in April that increasingly granular pricing could push firms toward charging close to the maximum each shopper will pay, a practice they describe as perfect price discrimination.
Key Conclusions
The first conclusion is that the technology is not the problem; the disclosure is. Dynamic pricing has existed for decades in airlines, hotels and ride-hailing, where customers accept that the price moves with demand. Grocery is different because shoppers treat shelf prices as stable reference points, and because a price that differs by shopper is invisible unless the retailer says so. The moment customers suspect that the price depends on them rather than on the product, the reference point collapses.
The second conclusion is that electronic shelf labels are a governance decision disguised as an operations upgrade. The stated benefits are real: fewer paper tags, faster updates, fewer price mismatches at the till. But the same infrastructure that lets a store fix a wrong tag in seconds also lets it reprice a category several times a day. Retailers adopting the hardware need a written rule about what they will and will not do with it, before the capability outruns the policy.
Where the legal risk sits
Several United States states are already moving. New York requires most businesses that use personal data to set prices to disclose it clearly, Maryland has restricted food retailers and delivery services from using personalised data-driven pricing to charge more for certain food, and New Jersey and Connecticut have enacted measures aimed at surveillance pricing. The pattern is consistent: regulators are less interested in the algorithm than in whether the customer can tell that a price is personalised.
Why grocers keep adopting it
Grocery margins are thin and waste is expensive. Marking down perishables as they approach expiry recovers value that would otherwise be written off, and doing it through an app or a digital tag is cheaper than manual relabelling. The operational case is strong. The reputational case depends entirely on whether the same system is used to raise prices for captive shoppers, which is the line regulators are drawing.
The Inflation Measurement Problem
Bank of England economists noted a statistical consequence that rarely appears in retail discussions. Consumer price indices are built from a representative sample of prices, which works when prices move slowly and uniformly. When prices shift continuously and differently for each shopper, the idea of a representative price becomes strained, and measured inflation diverges from what households actually experience.
That divergence matters commercially. If official inflation understates the experience of a household that pays more because of its browsing history, the household concludes that the statistics are wrong and that the retailer is responsible. Trust in the price signal erodes on both sides, and promotions lose their ability to communicate value because the reference price is no longer shared.
The comparison problem
Price comparison depends on a common reference. If two shoppers see different prices for the same item, neither can tell whether the difference reflects cost, demand or their own profile. Academic analysis cited by CNBC frames the risk plainly: when consumers cannot understand why they received a price, cannot compare it with others, and cannot switch easily, the disciplining effect of consumer choice weakens.
Best Practices
The first practice is to publish a pricing policy in plain language. A short statement covering what data is used, whether prices differ between channels, and how promotional prices are set does more for trust than a detailed privacy notice. Retailers that have published such statements report fewer complaints about price fairness, because customers can distinguish a markdown from a personalised quote.
The second practice is to separate the markdown engine from the pricing engine. Markdowns driven by expiry dates are defensible and easy to explain; prices that vary by customer profile are not. Keeping the two in different systems, with different approval paths, prevents an operational tool from quietly becoming a revenue tool.
Audit the tag, not just the price
Electronic shelf labels fail in specific ways: a tag that did not sync, a tag showing a unit price that does not match the pack size, a tag left on a discontinued item. These failures look like pricing errors to customers and like system errors to staff. A weekly audit of a random sample of tags, compared against the point of sale record, catches most of them before they become complaints.
Common Mistakes
The most common mistake is assuming that customers object to price changes. They do not; they object to unexplained differences. Airlines change prices hourly and face little backlash because the mechanism is understood. Grocers that change prices without explaining the mechanism inherit the suspicion that airlines avoided, and they inherit it in a category where shoppers visit weekly.
A second mistake is treating electronic shelf labels as a labour saving project only. The savings are real but modest, while the trust exposure is large. Retailers that budget for the hardware without budgeting for the policy, the audit and the customer communication capture the cost and the risk without the benefit.
What Regulators Will Ask Next
The next round of regulatory questions is likely to focus on evidence rather than intent. Regulators will ask retailers to demonstrate that a price difference between two customers is explained by a documented factor, such as a loyalty tier or a delivery cost, rather than by an inferred willingness to pay. Companies that can produce that documentation quickly will spend less time in disputes.
The practical preparation is unglamorous: log every pricing rule, record who approved it, and keep the version history. Most retailers already have this for promotional pricing because finance requires it. Extending the same discipline to algorithmic pricing is a process change, not a technology project.
Summary
AI pricing is arriving in grocery through a side door labelled operations. The same tags that fix a wrong price in seconds can reprice a category hourly, and the same data that improves forecasting can estimate what each shopper will pay. The technology is neutral; the policy is not.
Retailers that publish a clear pricing rule, separate markdowns from personalised pricing, and audit tags weekly will keep the operational gains without inheriting the trust cost. Those that skip the policy work will find that the cheapest efficiency gain in the store is also the most expensive one to explain.
Data Sources
1. CNBC analysis of AI dynamic pricing
2. Apparel outlook and tariff refunds
3. Target store operations app
FAQ
What is dynamic pricing?
A: Changing prices in response to market conditions such as demand, timing, capacity or competitor prices. Airlines and hotels have used it for years.
How is personalised pricing different?
A: Personalised pricing uses information about the individual shopper to estimate willingness to pay, rather than responding to market-wide conditions.
What does McDonald say about the lawsuit?
A: The company denies using AI to determine what individual customers will pay and says it provides franchisees with tools, research and recommendations.
Which states regulate this?
A: New York requires disclosure when personal data is used to set prices, Maryland restricts personalised pricing for certain food, and New Jersey and Connecticut target surveillance pricing.
What should retailers do first?
A: Publish a plain language pricing policy, keep markdown and personalised pricing in separate systems, and audit electronic shelf labels weekly against point of sale records.
