On September 29, 2026, Reuters reported that McDonald's uses an AI pricing engine to set what it calls the "optimal price" for menu items at individual US restaurants, according to Engadget's and Yahoo Finance's coverage. The story also appeared in the October 2 Hacker News AI discussion digests.
- How it works: the engine draws on transaction data from millions of orders across more than 13,000 stores, plus competitors' prices, to estimate each location's customers' willingness to pay. Reuters, citing two former employees, says the platform is run by Tiger Analytics, with McDonald's supplying the rules and targets.
- The effect: in Fresno, California, a Big Mac cost $5.69 in the app at one restaurant and $6.89 at a corporate-run location two miles away, a 21% difference.
- Franchisees: several told Reuters they are pressed to follow its recommendations and that departures are recorded. McDonald's called the reporting "speculative and uninformed" and the tool "a tool, not a mandate."
The timing is notable. Maryland's Protection From Predatory Pricing Act took effect on October 1. According to Skadden's summary, it bars food retailers and third-party delivery services from using a consumer's personal data to set higher food prices for that consumer, with penalties up to $10,000 per violation ($25,000 for repeat violations), enforced by the attorney general. "Food retailer" is defined as a store of at least 15,000 square feet selling tax-exempt food, and loyalty programs and published group discounts are exempt.
I think this story shows why the first wave of surveillance-pricing laws may miss the practice that actually spreads. Maryland's law is built around a person: their data, their profile, their higher price. On the reporting so far, McDonald's engine prices a place, using aggregate willingness to pay around a store, and a fast-food restaurant is unlikely to meet the 15,000-square-foot threshold anyway. So the most visible AI pricing system in the news this week probably sits outside the most visible law, as far as I can tell from the published text. That matters because location is a proxy. Pricing by neighbourhood can reproduce much of what pricing by person does, without touching anyone's personal data. The engineering lesson is that these systems optimise whatever target they are given, and "what this store's customers will bear" is a target, not a neutral fact. My expectation is that the next round of bills will move from regulating inputs (personal data) to regulating outcomes (unexplained price gaps for the same item), which is harder to write and harder to evade. Watch whether Maryland's attorney general or the FTC, which has already studied surveillance pricing, treats location-based AI pricing as in scope.
Reuters' account of McDonald's per-store AI pricing landed as Maryland's personal-data surveillance-pricing ban took effect, highlighting that store-level, location-based AI pricing may fall outside laws written around individual data.