American shoppers are being told that AI now sets their prices, and the evidence is more mixed than the headlines. The Associated Press reported on October 18, 2026 that consumer apprehension is growing as US retailers adopt AI-assisted pricing tools, citing Walmart, Whole Foods, Kroger, Target, Instacart, Amazon and McDonald's. This post separates three things that get blurred together: digital price tags, dynamic pricing, and personalized "surveillance" pricing.
Short version: the confirmed problems so far involve personalized or location-based pricing and legal fights about it, not shelf tags that change by the minute. We build on our earlier coverage of the McDonald's AI pricing lawsuit and of a study showing AI agents steering wealthier users to pricier options.
TL;DR: what is verified
| Claim | Status | Source |
|---|---|---|
| Digital shelf labels change prices constantly | Not supported. Retailers typically update weekly; changes during the day are "exceedingly rare" in grocery, per Relex Solutions' Asa Farquhar | AP |
| Walmart labels scan faces or phones | Described as false; CEO John Furner wrote "We price the product, not the person" | AP |
| Researchers saw surge pricing after labels arrived | An unnamed grocery chain study found "virtually no surge pricing" | AP |
| Target used app location to raise prices | Target agreed to a $5 million settlement and a ban on geofencing to raise prices; Target says its pricing is not personalized | AP |
| Instacart showed identical items at different prices | A Consumer Reports, Groundwork Collaborative and More Perfect Union investigation found it; Instacart ended tests that randomly assigned shoppers to groups | AP |
| Amazon used an algorithm to extract over $1 billion | An FTC allegation about "Project Nessie"; Amazon says it "grossly mischaracterize[s]" the tool, which it says was scrapped before the suit | AP |
| McDonald's coordinates prices with nonpublic competitor data | A consumer lawsuit allegation; McDonald's calls it "filled with inaccuracies" | AP |
Shelf labels, dynamic pricing, personalized pricing: three different things
A price tag riding a wave line, illustrating AI-assisted retail pricing that rises and falls with demand
Electronic shelf labels are screens that replace paper tags. Their main stated benefit, per Walmart's CEO, is saving workers time and keeping prices current. Unions oppose them and consumer advocates warn they could enable surge or personalized pricing later. That is a warning about future capability, not a documented practice.
Dynamic pricing is prices that move with supply and demand, the way Uber surge pricing raises fares when driver supply falls short, and the way airlines and hotels have long worked. It is legal and widely accepted in those sectors, and controversial in grocery and fast food.
Personalized or surveillance pricing tailors the price or discount to an individual: location, income, browsing history. This is where regulators and lawsuits are concentrated. The AP also stresses that different prices do not automatically mean personal data was used.
What does the data say about how common AI pricing is?
The AP cites a Simon-Kucher survey of more than 2,200 business leaders in 28 countries across 39 industries: 41 percent use AI to gather market intelligence, including competitors' prices, and 28 percent use it for "real-time price optimization." That is a global, cross-industry survey of business leaders, not a count of US supermarkets repricing items, so it does not show how often individual shoppers see different prices.
Decodo, a web data company, told the AP its 2025 tracking found about half of US price changes were decreases. The National Retail Federation's position, per the AP, is that algorithmic pricing is a new tool for practices retailers have always used. Kearney's Katie Thomas described "major tensions happening right now between who's really going to have the power" over algorithms.
The FTC's January 2025 staff study, Surveillance Pricing 6(b) research summaries, reviewed six intermediary firms and found they worked with at least 250 clients, including grocery and clothing retailers. It indicated a wide range of personal data, such as location, browsing and mouse movements, could feed individualized prices. The report used hypothetical examples, was approved 3-2, and was an interim look rather than a final report.
The lawsuits and rules, in plain terms
- Target: California prosecutors sued in 2022, alleging its app showed lower prices before shoppers entered stores and higher prices inside. Settlement: $5 million and a ban on using geofencing to raise prices.
- Amazon: The FTC's 2023 antitrust suit targets "Project Nessie," which the agency alleges produced more than $1 billion in excess profits. A trial is scheduled for next year, per the AP.
- McDonald's: A recent consumer suit alleges its pricing system uses nonpublic sales data from competing locations. The tool is offered to US franchisees, and the company disputes the claims.
- States: Maryland and New Jersey recently restricted using personal data to set grocery prices, with exceptions for some loyalty discounts and promotions. Connecticut covers a broader range of purchases. At least 30 states considered pricing bills this year, per the National Conference of State Legislatures.
- FTC: It proposed an enforcement policy saying failure to disclose personal-data use in individualized pricing could violate consumer protection law.
Why the mix-up keeps happening
Part of the confusion is timing. Walmart's plan to put electronic shelf labels in all US stores arrived in the same news cycle as lawsuits over pricing algorithms, so viral posts fused them into one story. The AP notes that social media posts falsely claim the Walmart labels scan faces or phones. The company's CEO answered in a blog post the previous month, and the AP's reporting supports a narrower reading: a label is a display, and a price is a decision made somewhere else in the system.
That distinction matters for how you judge any future claim. A display that can change prices quickly is a capability. A pricing model that uses your location or browsing history is a practice. A lawsuit alleging such a practice is an allegation until a court or settlement says otherwise. Target's settlement is the clearest example in the AP's account of a practice that ended in a legal consequence, and even there the company says its pricing is not personalized.
The second source of confusion is that "AI" now labels everything from a simple rule engine to a learned optimizer. A retailer that raises a price on a rainy day with a spreadsheet rule and a retailer that uses a trained model are doing different things technically, and may be doing the same thing legally. State laws tend to focus on the inputs, personal data, rather than on whether the method is called AI, which is why the Maryland, New Jersey and Connecticut measures matter to anyone building pricing or shopping software regardless of the technology they use.
Where do AI agents fit in?
Two threads make this an AI story and not only a retail one. First, retailers are adding machine-learning price optimizers, which is the "AI-assisted" part. Second, shoppers are starting to hand purchases to AI agents, and our coverage found worrying signals there: the Et Tu, Brute study reported agents recommending pricier options to users they inferred were wealthy.
Stores are also deciding who gets to shop on a customer's behalf. We covered Amazon blocking agents and what Paul Graham called a startup opportunity, Shopify's move to let any AI agent shop, and Amex's framework for agents making unauthorized purchases. Banks are drawing similar lines: see Citi, JPMorgan and Bank of America blocking agents. If both the store and the buyer use algorithms, who sees which price becomes harder to audit.
What developers on Hacker News said
The AP story drew a small Hacker News thread (about six points when we checked), so treat this as anecdote. One commenter, iugtmkbdfil834, said the topic is "fascinating" and more tractable than claims that chatbots steer people, because some stores already did this. Another, silverFork, argued that merchants can know a visitor's worth in real time through ad data matched to IP addresses, and that AI could become a new consumer of that data. Both are opinions from commenters, not verified facts.
What should shoppers and builders do?
A small cart with a price tag, a reminder to compare prices across app, web and store
Shoppers: Compare the same item in the app, logged out on the web, and in the store. Screenshot both with timestamps. Check whether your state has a disclosure or personal-data pricing law, and whether the retailer discloses data use. A price gap alone is not proof of personalization.
Builders of shopping agents: Log the price shown, the context and the user profile signals sent to the retailer. If an agent recommends an option, be able to explain why. The law is moving toward disclosure of personal-data use, so keep an audit trail now.
Teams deploying pricing models: Document which inputs the model uses. Exclude personal and location-linked attributes unless legal has signed off, and test for disparate outcomes before launch. Agent actions that touch payments or accounts deserve guardrails and human approval.
What to watch next
- The Amazon Project Nessie trial next year, which will test how courts treat pricing algorithms.
- Whether more states copy Maryland, New Jersey and Connecticut, and whether the FTC finalizes its proposed enforcement policy.
- Whether independent researchers publish new price-tracking studies for Walmart's planned rollout of electronic shelf labels across US stores.
- Outcomes in the McDonald's case and any similar suits over shared pricing tools.
We will update this post if the facts change. We could not read the AP original on its own site, so quotes and figures here come from the syndicated copy on ABC News and the FTC's own release; the Hacker News thread is extra context, not a source for the figures.
Related reading
- McDonald's AI pricing engine: what is alleged and denied
- Study: AI agents steer wealthier users to pricier options
- Paul Graham on Amazon banning agents
- Shopify lets any AI agent shop
- Amex framework for unauthorized agent purchases
- Banks block AI agents from customer accounts
- Official: FTC surveillance pricing study release
Facts reflect reporting available on October 18, 2026 and may change as cases and laws develop.