McDonald's AI priced the Big Mac. A court will now decide whether that was coordination

What the McDonald's lawsuit alleges, in plain terms#

On 2 October 2026 Michael Thomas, a McDonald's customer in Illinois, filed a proposed nationwide class action in the US District Court for the Northern District of Illinois, case number 1:26-cv-12149. The complaint runs to 31 pages and names one mechanism: a machine-learning system that, the filing says, has generated restaurant-by-restaurant price recommendations since at least 2019. Reuters, via the Claims Journal reports the suit accuses the chain of conspiring with its independent franchisees to fix menu prices using algorithms trained on nonpublic data.

The legal engine is Section 1 of the Sherman Antitrust Act, which forbids contracts and conspiracies in restraint of trade. The complaint's own line is short and hard to quarrel with on its face: “Independent businesses must set their prices independently”. The allegation is that they did not. Wide Open Country quotes the filing's account of the mechanism, that the algorithm carries rules tying one restaurant's price increases to whether other McDonald's restaurants have already raised prices. Read that sentence twice: it describes a shared signal, not a shared menu.

Thomas asks the court to certify a class of customers nationwide and to award damages. Certification is a long way off, and the company has already answered in public. The New York Post reported the filing on 5 October; McDonald's responded the same day, and we treat both positions as claims rather than findings.

Here is the structural fact that makes the allegation possible, and it is not in dispute. McDonald's owns and operates very few of its American restaurants. In a 2024 open letter, the president of McDonald's USA, Joe Erlinger, put the figure at more than 95 per cent franchised. The complaint's point follows from that spread: a recommendation engine reaching into that network is not one company setting its own prices alone, but one company shaping the pricing options of thousands of businesses that share its brand.

The vixen's first note in the margin: a fox that hunts alone is a fox; a fox that teaches the whole thicket to hunt in formation is something else, and the complaint has a word for it. That word, price fixing, is the next thing to get straight, and the distance between price fixing and dynamic pricing is the distinction at the centre of it.

What dynamic pricing is, and why McDonald's says it is not the issue#

Dynamic pricing means charging different prices for the same thing at different times, usually in response to demand. Airlines do it, hotels do it, and most ride-hailing apps do it while you watch the little car crawl towards you. It is legal, it is everywhere, and it is the reason dynamic pricing shows up in the same breath as this lawsuit, so the phrase deserves pinning down before it does any more work.

The distinction that matters is between three things people keep mixing. Dynamic pricing moves a price with demand at the moment of sale. Algorithmic pricing uses a model to decide what a price should be, at whatever cadence the business chooses. Price fixing is competitors agreeing, directly or through a shared mechanism, to stop competing on price. Only the third is unlawful, and the complaint's case is that this is what happened.

McDonald's rejects the framing outright. Its own statement, headed “Separating Fact from Fiction: AI Does Not Set Prices at McDonald's”, says the pricing portal provides restaurant-specific recommendations, does not set or change prices, and leaves the decision to franchisees. The document adds four words that matter here: “McDonald's does not use dynamic pricing”. The company told Reuters that its portal is a tool rather than a mandate, and that the use of pricing recommendation tools and analytics is widespread across industries.

That defence is not a dodge. It is a precise legal position. If the tool only advises, and each franchisee chooses, then the defence's account is that no agreement exists for Section 1 to reach. The plaintiff's answer is that a tool can be optional in its terms and compulsory in its consequences, and that a company which logs who deviates and discusses non-compliance in business reviews has built something that functions like an agreement. McDonald's says it takes antitrust compliance seriously and that the warnings in its own terms of service are not evidence of anticompetitive behaviour.

Is dynamic pricing illegal?#

No, and this is the flattest answer in the piece. Dynamic pricing is ordinary commercial practice. Money magazine's survey of the state bills sets out the split cleanly: dynamic pricing is accepted for airlines and ride-hailing, while applying it to groceries and other essentials is where legislators draw a line. How far that holds depends on the jurisdiction, and this piece is scoped to the United States. Dynamic pricing becomes a legal problem when it stops being each business's own decision.

So the honest reading of the McDonald's defence is that it is fighting on the right ground, and the honest reading of the plaintiff's case is that the ground is narrower than the headlines suggest. Everything turns on whether an advice tool, plus monitoring, plus a business standard, adds up to an agreement. Whether dynamic pricing arrives by algorithm or by a manager's spreadsheet does not change that question. For a marketer, this is a story about data governance wearing a lawyer's robe, and our AI consultancy work starts from the same question: what does your own tooling actually do, and who can see it do it?

The Fresno Big Macs: $5.69 and $6.89, two miles apart#

The detail that gave the story legs is not a legal one. In September 2026, Reuters checked the McDonald's app and found a company-operated restaurant in Fresno, California, selling a Big Mac for $5.69, while another company-run location about two miles away charged $6.89. Same brand, same sandwich, roughly a fifth more, and Reuters was careful to say it could not confirm that the algorithm caused the gap, and nobody has shown whether the engine produced it at all. That is the heart of the dynamic pricing row.

Why two nearby restaurants can land on different numbers is where the engine's design matters. Reuters reviewed screenshots of the franchisee interface taken in August and described a system weighing local price sensitivity, an estimate of what customers in that area appear willing to pay, and competitors' public online prices. One message in the interface told an operator their restaurant was showing medium sensitivity to price. The trade coverage notes that the same engine pulls public menu prices from nearby Wendy's and Burger King locations. Both chains told Reuters they do not use AI to set prices, and both appear in McDonald's system only as public reference points.

We have elided the last word of that post because it is an expletive; the sentence ends there and the poster's next sentence hopes a retailer is next. It is here as one user's reaction, and because it shows the gap between how the case is argued and how it is read. The complaint says data and discretion. The timeline says a fox in the henhouse. Neither is evidence, and both are how the story travels.

This is the part a brand team should sit with. The most visible detail in the whole affair is not an allegation of collusion. It is two numbers two miles apart, which any customer can check for themselves in an app. Livemint's summary of the Reuters work makes the point that a pricing engine trained to find each neighbourhood's ceiling can produce exactly this kind of visible spread. The dynamic pricing argument is, in the end, an argument about visible arithmetic: customers can see the spread, and they can screenshot it. If you run anything similar on your own site, you have a communications exposure that a legal review alone will not close.

Inside the engine: Dynamic Yield, Tiger Analytics, and a quiet warning#

The tool did not arrive yesterday. McDonald's bought the personalisation firm Dynamic Yield in March 2019 for about $300 million, as Restaurant Dive reported at the time, and rolled machine-learning menu boards into thousands of US drive-thrus within months, choosing what to show by time of day and by what is popular at that restaurant. That acquisition sits at the start of the timeline the complaint points to, which is why the 2019 start date is not a rounding error.

Two people who used to work at the data consultancy Tiger Analytics told Reuters that the firm runs the engine, with McDonald's setting the business rules. Tiger declined to comment. The rules as described are worth reading twice, because they are less about any single price than about how increases propagate: lean on items that have not moved in two years, avoid raising ice cream and drinks in summer, and concentrate on categories where at least 30 per cent of restaurants have already gone up. That last rule is the one a competition lawyer circles in red ink, because it makes one restaurant's increase a reason for the next.

CNBC reported the franchise standard that took effect on 1 January 2026, under which McDonald's assesses whether franchisees are delivering value for customers. Reuters put the sharper version of it: a June franchisee document records deviations from the recommended prices in detail, and the CEO, Chris Kempczinski, told investors in August that pricing non-compliance in certain cases is part of those business-review conversations.

One former franchisee told Reuters she never felt compelled to take the suggested prices, and five others described pressure of various kinds. Karen King, a former store owner, put it plainly: that you do not really have much of a choice any more. McDonald's did not address her specific claims. The company did say that its terms of service carry an antitrust warning, and that the warning is not evidence of wrongdoing. Here is the fox's reading of that signal. If a platform's own documentation carries an antitrust warning, that is worth reading closely. It is not by itself an admission of anything, and it is not by itself a defence either.

Strip the brands away and this is a story about AI pricing as a category, or about dynamic pricing if you prefer the older label. Every vendor selling algorithmic pricing, and every brand buying it, now has a worked example of what a filing will quote back: the model's rules, the deviation log, the performance conversation. That is a governance problem long before it is a legal one, and it is why the practical material below is aimed at the operator rather than the litigator.

Who else is exposed: RealPage, 26 states, and a rule taking shape#

McDonald's is not the first defendant to face this argument, and the earlier case is a better guide than the headlines. In November 2025 the US Department of Justice settled with the rental software firm RealPage over claims that its revenue management tools helped landlords align rents. Fenwick's antitrust team read the settlement as a blueprint.

The rules it sets out are precise: no nonpublic, competitively sensitive data from competing landlords in the recommendations a tool produces at the moment of pricing; models trained only on backward-looking data; and product design that does not default towards increases. Those rules belong to one agreed settlement rather than a general safe harbour, and they are the clearest published guidance on algorithmic pricing so far.

On this, state legislatures are the ones writing rules. Money magazine's May 2026 survey counted more than 50 bills across 26 states aimed at restricting or banning what campaigners call surveillance pricing, with Maryland signing the first into law and New York's Algorithmic Pricing Disclosure Act as the model that others copy. That fight is not a referendum on dynamic pricing in general; it is a line drawn around food and other essentials. PYMNTS reported the trend and the counter-argument, that dynamic pricing is already normal in travel and can work in a customer's favour. Both things are true, and the essential-goods line is where the legislative attention sits.

The consumer evidence is thinner than the politics. A Consumer Reports investigation found prices for identical products varying by as much as 23 per cent in tests across grocery chains, and estimated a cost of up to $1,200 a year for a family of four. That study looked at personalised, app-driven pricing rather than restaurant menus, so treat the figure as context for the debate rather than evidence about McDonald's. We include it as context, with the sources named.

Put the three layers together and the direction is clear. A settlement in housing, a wave of state bills aimed at retail, and a class action in fast food all describe the same line: a pricing algorithm may look outward at public information, and may learn from a company's own history, but it should not launder one competitor's confidential numbers into another's price. A brand that can draw that line inside its own stack is in a stronger position. A brand that cannot may find the question asked by a customer, a partner or a regulator. That is the practical shape of the dynamic pricing argument.

The Monday plan for brands running their own pricing tools#

Most folkfox readers are not McDonald's and never will be. Plenty of them do run pricing tools, revenue dashboards or dynamic offers on their own sites, which is why the useful thing to take from this case is a checklist rather than an opinion. None of it turns on whether dynamic pricing is wise or fair, only on the plumbing behind it. An afternoon with the right five questions is a reasonable starting point.

For every pricing tool we run, list:
1. What data goes in, and whose it is (ours, a vendor's other clients, public).
2. Whether any nonpublic figure from another business enters a live price decision.
3. Who can see the output, and who approves a change.
4. What we log, and for how long.
5. What we would hand an auditor if asked why a price moved on a given day.

Three questions for any pricing vendor:
1. Does your live recommendation use any nonpublic data from your other customers?
2. Does the interface default towards increases, and can we turn that off?
3. What is your own documented position on antitrust compliance, and can we see it?

Two rules of thumb make that list easier to hold in your head. First, public in, private out: a tool may ingest what anybody could see, and may learn from your own history, but it should never carry another business's confidential numbers into your decision. Second, the human stays visible: a named person approves, and the record says who. Those two habits also happen to be exactly what good analytics governance has always asked for, which is the reassuring part.

None of this is legal advice, and nobody knows yet how a court will read an advice tool plus a monitoring log. What we can say with confidence is narrower and more useful. The exposure in these systems sits less in the model than in the seams around it: the shared data feed, the deviation report, the performance conversation. Our SEO and GEO work, our content practice and our brand work all sit downstream of those seams, because a brand that cannot explain its own numbers ends up explaining them to somebody else.

The fox does not wait in the open field. It studies the hedgerow, the gate and the gap under the fence, and then it moves. A pricing lawsuit is the open field. Your data flows, your logs and your approvals are the hedgerow, and they are the part of this story closest to your control.

Frequently asked questions#