A federal class action filed in Chicago alleges that McDonald’s is coordinating menu prices across its restaurants through an AI-powered pricing engine — conduct the suit says breaches US antitrust law — a system that, according to Reuters reporting picked up by CNBC, continually analyses data from millions of daily transactions across nearly 14,000 restaurants. McDonald’s denies the allegation outright: “AI does not set the price of a Big Mac or any other menu item,” the company says, insisting franchisees make their own pricing decisions and that analytics tools are widespread across industries. The claim is untested — but the lawsuit has landed on the exact week that the infrastructure for algorithmic food pricing became impossible to ignore in ordinary shops.
What the lawsuit actually alleges
Per Claims Journal’s summary of the Chicago filing, the proposed nationwide class action alleges the fast-food company coordinates menu prices across franchises and company-owned stores “through an AI-powered pricing system.” Reuters reported the pricing engine uses machine-learning algorithms fed by transaction data. The company’s formal response is that it provides franchisees “tools, resources, research and recommendations to help them make informed decisions” — recommendation, not determination.
That distinction matters legally and practically. McDonald’s restaurants are overwhelmingly franchisee-owned; corporate can recommend, but the franchisee signs the price change. Whether recommending with national-level data crosses into coordinating is the antitrust question the suit raises, and it mirrors a wave of US class actions alleging algorithmic price coordination in hotel rooms and apartment rentals. Nothing has been proven; McDonald’s position deserves to be stated alongside the claim, and here it is.
The quieter story: the shelf is already digital
While the lawyers argue about Big Macs, CNBC’s survey of food retail catalogues how far the enabling technology has already spread. Kroger uses an AI platform called FlashFood to mark down perishables approaching the end of shelf life and push those offers through an app. Electronic shelf labels — digital price displays that can be updated centrally, instantly, store-wide — are rolling out at Kroger, Amazon Fresh, Walmart and Whole Foods in the US, and at Tesco, Morrisons and Asda in the UK. Once every shelf label is a screen, the cost of changing a price drops to zero, and zero-cost repricing is the precondition for dynamic pricing.
Bank of England economists Clare Lombardelli and Rupal Patel, writing in an official BoE analysis of personalised pricing, document how far this has already gone in travel: the share of UK hotel rates changing at least monthly has risen from about 15 percent in 2005 to roughly 80 percent today. They define the two tiers precisely — dynamic pricing changes prices with market conditions; personalised pricing goes further and tailors the price “to their personal circumstances and consumption patterns.” Their conclusion cuts both ways: so far the shift has not pushed inflation systematically up or down, and outcome depends on competition and information availability. University of East London commercial-law lecturer Miroslava Marinova, speaking to CNBC, points out dynamic pricing itself is not new — airlines, hotels and ride-hailing have done it for years.
Why a Big Mac lawsuit is the test case
Food is the last category where consumers still expect the price on the shelf to be the price for everyone. Airlines trained customers that the person beside them paid a different fare; nobody has trained them for it at the supermarket. The McDonald’s suit matters less for its merits — which a court will decide — than for the line it draws: if recommending prices with national transaction data can be alleged as coordination, then every retail AI platform feeding pricing signals to thousands of independent stores is inside the same doctrinal zone.
The other lesson of the filing is structural. The alleged mechanism runs through franchisees — independent businesses receiving central recommendations generated from pooled data. Grocery is built the same way: chains centralise merchandising while stores execute, and vendor platforms increasingly sit in between. An AI that nudges thousands of nominally independent price-setters in the same direction at the same time is functionally similar whether it lives at McDonald’s corporate or at a shelf-label vendor serving fifty chains.
Our take
New Zealand is already one step down this road: Countdown’s owner Woolworths NZ was fined in 2024 for shelf-price mismatches, a scandal that was fundamentally about the gap between displayed and charged prices — and electronic shelf labels eliminate mismatch by making display and charge the same database entry. That cuts two ways. ESLs make shelf-jumping errors disappear, but they also remove the natural friction that stopped a price from changing while you walked to the till, and they make surge-style repricing physically possible in a country whose consumer law still assumes prices are stable enough to compare.
The realistic near-term NZ experience will not be movie-ticket-style surge pricing on milk. It will be app-mediated: personalised markdowns in the loyalty app (already happening at both major chains’ parent companies), algorithmic clearance pricing on short-dated stock, and stable shelf prices that quietly hide a spread of personalised deals underneath. The BoE’s framing is the honest one — whether that world is bad for shoppers depends on competition. NZ grocery is famously not competitive: two companies control the market, and the Commerce Commission’s 2022 study found prices were persistently above competitive levels. Dynamic pricing infrastructure inside an uncompetitive market is not a neutral bet; the same screen that shows you a markdown can show the duopoly your willingness to pay.
A weekly shop’s worth of related coverage: the AI price divide that already splits who gets the best models, NZ electronics prices surging as AI devours memory chips, and Walmart’s warehouse robots fighting peak complexity — the supply end of the same algorithmic stack.
