Instacart has closed an experiment that used artificial intelligence to set bargaining prices for grocery items amid concern the practice was misleading customers, according to a study from its customers. The move comes amid increasing regulatory scrutiny and a tide of public criticism over opaque pricing practices in digital grocery shopping.
Why Instacart Yanked the Cord on A.I. Pricing Tests
The move followed a multicity study of 437 volunteers by Consumer Reports, Groundwork Collaborative and the More Perfect Union, which simulated the same shopping trips. Some 74% of participants noted price differences for identical products, with an average spread between highest and lowest prices in test results of 13%. One estimate from the researchers said that prolonged fluctuations in prices could add more than $1,200 to annual food costs of a typical family of four, an amount Instacart challenged.
Instacart said the tests were meant to determine price sensitivity, not to individualize prices according to who a shopper is. The company noted the experiments were randomized A/B tests run at retailers’ request and enabled by Eversight, an AI tool Instacart bought in 2022 to help brands and stores test promotions and pricing. Still, the practical result — shoppers seeing different prices for the same item — undermined trust.
What the A/B Pricing Tests Did, in Fact and Practice
In classic A/B testing, for instance, two sets of people are shown different price points to measure demand and identify a price that will increase sales or margin. It’s a standard practice in e-commerce and has been for some time offline through regional pricing and in-store promotions. The controversy here was about unclear labeling, and the idea that two people in the same city shopping at the same time could be shown different prices on a platform they weren’t aware was testing its software.
Instacart now says that it will display uniform prices to all shoppers for the same items from the same store at a given time. In practice this means if two households make the exact same cart for an order from the same store on Instacart, they should no longer see bottom-line differences due to experimentation.
How Regulators and Lawmakers Responded to the Tests
Regulatory scrutiny intensified quickly. Meanwhile, on Monday night Reuters was first to report that the Federal Trade Commission had issued Instacart a civil investigative demand for information about how a company called Eversight is being deployed in pricing experiments. An agency spokesman publicly expressed consternation about what it had read in the press. In a separate letter, several members of Congress called on the F.T.C. to investigate whether consumers were being intentionally kept in the dark about how they were seeing prices on Amazon and suggested that the commission require on-screen disclosures whenever price experiments are live.
Regulators have been paying closer attention as algorithmic pricing and “dark patterns” — tricks used online to get consumers to make purchases they ordinarily wouldn’t — come under broader scrutiny. A/B testing alone is not illegal, but undisclosed practices that change price or choice in a meaningful way can attract attention under Section 5 of the FTC Act. Critics say that hidden experiments threaten to undermine pricing transparency at a time when families are still wary of spiraling food costs.
What Changes for Shoppers on Instacart Going Forward
Instacart has said it ended all item-level price tests on its marketplace and turned off Eversight-powered experiments for retailers using the platform. The company says retailers still determine their own prices, but any testing that introduces side-to-side variance among shoppers on Instacart is out the window. The policy seeks to ensure, for example, that two customers standing a mile apart don’t see different price tags on the same cereal box or milk jug.
The reversal also is part of a wider trend by platforms to cut back on fees and pricing. Instacart recently settled an F.T.C. lawsuit involving how it characterized service fees, and agreed to refund $60 million to customers that were affected. In totality, these actions appear to indicate a recalibration towards simpler, more predictable checkout processes.
The Bigger A.I. Pricing Debate and What It Signals
Algorithmic pricing is now the de facto norm across a slew of industries, from airline seats to rideshares to hotel rooms, so it was only a matter of time before this form of market-driven automation took over grocery delivery.
The important thing to regulators and customers, the distinction that would matter most, is transparency. Dynamic pricing that reacts to supply-and-demand signals is one thing; blind experimentation that results in different prices for near-identical shoppers is another, particularly in essential categories like food and household staples.
Retail analytics experts said testing can help fine-tune promotions and cut waste by more tightly tailoring a store’s prices to local demand. But in experiments left without visible guardrails, they can veer into the territory of being seen as unfair or discriminatory. Clear notice-and-disclosure regimes, pathways to opting out and framework conditions — say, ensuring a level playing field among buyers over a period of time within a store — now seem like useful baselines for regulation.
Instacart’s withdrawal and other such controversies highlight a new reality for AI in commerce: that even if a practice is legal and data-driven, it also has to feel fair. For now, the company says everything in that same basket at that same moment will cost everyone the same — a simple promise that shoppers — and regulators — can readily test.
