Did McDonald's hand Big Mac pricing to a machine-learning model? A September 29 Reuters investigation says the company runs an algorithmic engine that proposes an "optimal price" for each menu item at each of its restaurants. Within days, a customer filed a nationwide antitrust class action, and McDonald's published a rebuttal saying AI does not set prices. The Hacker News thread on the story reached about 60 points. This post lays out what is reported, what is alleged, what is denied and what nobody has shown yet.
We could not open the Reuters article itself (it sits behind a login wall), so the details of the investigation below come from outlets summarizing it, such as TechSpot and Newsweek, plus McDonald's own statement. Treat numbers as Reuters' reporting, relayed.
TL;DR: reported, alleged, denied
| Question | Short answer |
|---|---|
| What did Reuters report? | A machine-learning engine analyzes transaction data from nearly 14,000 restaurants and recommends an optimal price per item per location |
| Does it use competitor prices? | Reportedly yes, including menu prices at Wendy's and Burger King |
| Is there a willingness-to-pay factor? | Franchisee screenshots reportedly show labels like "MEDIUM SENSITIVITY to Price," based partly on customer willingness to pay in the area |
| What is the lawsuit? | A nationwide class action filed October 2 in federal court in Chicago, alleging price-fixing through a shared algorithm |
| What does McDonald's say? | AI does not set prices; the portal is "a tool, not a mandate"; franchisees decide; no dynamic pricing |
| Is anything proven? | No. Reporting and allegations only, no ruling |
What Reuters reported
According to the secondary summaries, the system proposes a price for every item at every restaurant. It reportedly weighs local transaction history and competitor menu prices, and one factor is an estimate of how much a store's patrons will pay. Franchisee screenshots reportedly show prompts such as "Your restaurant is showing MEDIUM SENSITIVITY to Price."
The most quoted example is a Big Mac in Fresno, California: $5.69 at one company-run restaurant and $6.89 at another about two miles away, a 21 percent gap. Reuters reportedly could not confirm whether the engine's recommendations or other factors, such as local costs, produced the difference. That caveat matters: a price gap on its own does not show a model caused it.
The investigation also reportedly says five franchisees felt pressure to use the tools. Franchisees are independent operators, and McDonald's says they set their own prices. Reuters reportedly cites an internal communication from January requiring franchisees to be "constructively engaging with McDonald's approved Pricing Consultant and Tools." Whether that language amounts to pressure or ordinary compliance with a brand program is exactly what the lawsuit will argue over.
Reuters also reportedly asked competitors: Wendy's and Burger King said they do not use AI in pricing decisions. Wendy's drew backlash in 2024 over its CEO's remarks about dynamic pricing and has said those comments were misconstrued.
Retail price tag on a shopping bag illustrating McDonald's AI pricing recommendations at restaurant level
What the class action alleges
Gavel beside a scale representing the McDonald's AI pricing class action and its algorithmic pricing antitrust claims
Per Newsweek, the complaint was filed October 2 in federal court in Chicago by plaintiff Michael Thomas, who seeks to represent potentially millions of customers. The claim is that McDonald's violated U.S. antitrust law by conspiring with independent franchisees to fix prices. The mechanism alleged is an algorithm trained on nonpublic pricing data that produces recommendations across U.S. restaurants. The core theory: franchisees should price independently, and a common algorithm fed pooled, nonpublic competitor data suppresses that competition. The complaint cites the Reuters investigation.
This is the same legal shape as earlier algorithmic-pricing disputes, where plaintiffs argue that several nominally independent sellers using one vendor's recommendations behaves like a cartel. It is also adjacent to the allegations in our coverage of the AI model-slowdown cartel lawsuit, where the theory is coordination among rivals rather than a shared tool. Different facts, similar question: when does common software become common conduct?
What McDonald's says
On October 1, McDonald's published Separating Fact from Fiction: AI Does Not Set Prices at McDonald's. Its main points:
- AI does not set the price of a Big Mac or any other menu item, and does not determine what an individual customer pays.
- McDonald's does not use dynamic pricing.
- Franchisees independently set menu prices at their restaurants.
- The pricing tool gives restaurant-specific recommendations based on local market dynamics; people make the final decisions.
In a statement to Newsweek on October 6, the company called the claims "speculative and uninformed" and said the portal is "a tool, not a mandate." It denies the allegations and says the suit tries to recast "a standard business practice as something controversial."
Notice that the Reuters description and the company's statement are not necessarily contradictory. A system can use machine learning to recommend prices while the company insists humans decide. The disagreement is over characterization, and over how much freedom a franchisee practically has when the brand requires engagement with the tool.
Recommending versus deciding: why the wording matters
For anyone who builds or buys AI, this is a familiar problem. A model that outputs a recommended price is decision support. It becomes an automated decision when no one realistically overrides it, or when the business rewards following it. Regulators and courts tend to look at practice, not at the label on the dashboard.
Three practical questions separate the two:
- Defaults. Is the recommended price pre-filled, or does the operator type one in?
- Friction and incentives. Is deviating from the recommendation tracked, questioned or penalized?
- Inputs. Does the model train on pooled data from competing operators that those operators could not see individually?
The third is the heart of most algorithmic-pricing cases. Pooling nonpublic data across supposedly independent sellers is what turns a forecasting tool into an alleged coordination device. If you are building anything similar for a multi-tenant product, that is the design question to document early.
Spreadsheet grid with one highlighted cell representing a recommended menu price per restaurant
Restaurant-level pricing versus personalized pricing
Many readers will hear "AI sets your price" and assume the system quotes different prices to different people at the same counter. Nothing reported so far says that. The engine, as described, works per restaurant: it estimates local sensitivity and proposes one menu price for everyone in that store. McDonald's says it does not tailor prices to individual customers or change them in real time.
That distinction is important because the research on individualized pricing is a separate topic. Our write-up of the Et Tu, Brute? paper covers simulated tests where personal AI agents steered users they inferred to be wealthy toward pricier options. Those are recommendations from a user's own agent, not store prices from a retailer's model. The McDonald's story is about the retailer side. Both feed the same public worry: that models learn how much people can bear to pay.
What developers on Hacker News are saying
The Hacker News thread has about 60 points and 38 comments. These are individual commenters' views, not verified facts.
- One commenter, SMAAART, asked how willingness-to-pay pricing differs from the opaque pricing already common in enterprise SaaS and consulting, a fair point about how normal the practice is outside consumer food.
- tecoholic suggested McDonald's pricing might survive scrutiny because the engine appears to charge the same price to everyone at a given store.
- lurkingllama reported seeing a Big Mac meal at $10.80 at one location and a Big Mac at $8 at another, anecdotal evidence of the spread, not proof of cause.
- grogers argued the company should fix its slow ordering kiosks before spending on price optimization.
Many other comments were jokes or boycotts, which we skip.
What this means for what you build or pay
- If you ship pricing or ranking models: record whether the human can and does override, and avoid training on competitor data that the competitor's users cannot see. The legal exposure follows the practice.
- If you are a franchisee or small operator: read your vendor's contract for language about required use of tools. "Constructively engaging" can read very differently to a lawyer than to a manager.
- If you are a customer: the same sandwich costing different amounts at nearby stores is not new, and by itself does not prove an algorithm caused it. Checking the in-app price versus the counter price is the practical test.
- If you track AI regulation: this joins a growing set of cases where an ordinary business tool is reframed as automated decision-making. Our AI lawsuit tracker shows how fast that docket is growing, and the consumer spending data from a16z shows how little of the public actually pays for AI directly while meeting it through products like this one.
What we do not know
- Whether the engine's recommendations caused the Fresno gap or other price differences.
- How often franchisees follow the recommendations, or what happens when they do not.
- Whether the complaint survives a motion to dismiss. Nothing has been ruled on.
- What the training data actually contains, and whether any of it is nonpublic competitor data in the sense the complaint alleges.
- Whether regulators will open their own inquiry. None has been reported in the sources we reviewed.
How to follow the case
Watch for McDonald's response in court (a motion to dismiss is the usual first step), any filings that describe the tool in technical detail, and any regulator statements. We will add an update to this post if the facts change. For related reading on lawyers and AI in the courtroom itself, see our coverage of the $999.99 fine for AI fake citations, and for another test of AI agents in marketplaces, Anthropic's Project Deal.
Related reading
- Et Tu, Brute? AI agents steer wealthier users to pricier options
- AI model-slowdown cartel lawsuit
- AI lawsuit tracker: 248 cases
- a16z top 100 consumer AI spending
- AI agent privacy promises
- Anthropic Project Deal
- Reuters: Inside McDonald's push to have AI price your Big Mac
Details are accurate as of October 12, 2026. The lawsuit is at an early stage and all claims in it are allegations.