Instacart is shutting down an artificial intelligence pricing experiment that quietly showed different grocery prices to different users, a test that had been running inside its marketplace and rattling some of its most loyal shoppers. The company is now promising a return to simpler, more predictable pricing, even as it defends the program as misunderstood and insists it never let retailers charge more than the prices they set in stores.
The reversal lands at a delicate moment for the grocery delivery giant, which is trying to convince customers, regulators, and investors that it can innovate with AI without turning everyday essentials into a high-tech shell game. I see the decision as a revealing case study in how quickly consumer trust can evaporate when algorithmic experiments touch something as basic as the cost of milk, cereal, and baby formula.
Instacart’s AI pricing pivot, explained
Instacart has framed its move as a clean break from a controversial experiment rather than a retreat from technology altogether. In a corporate blog post, the company said it is ending “item price tests” that let some retailers use AI tools to vary prices on specific products for different customers inside the Instacart app, while stressing that the underlying store prices were still set by the grocers themselves. The company argued that the tests were designed to help retailers find lower prices that would appeal to shoppers, not to quietly ratchet prices upward.
At the same time, Instacart acknowledged that the program had become a lightning rod, saying that “much of” the public conversation around the tests was driven by misconceptions about what the company does and does not do with prices. In that same message, Instacart pledged to keep “clearly labeling” when prices in the app differ from what customers would see on store shelves so that people can understand and compare the costs they are being offered, a promise laid out in its update on ending item price tests.
How the Eversight engine powered the tests
The now-shelved experiments were built on technology Instacart picked up when it acquired Eversight, an AI-enabled pricing and promotions platform, in 2022. Eversight’s software was designed to let retailers run rapid-fire tests of different price points and discounts, then use machine learning to identify which combinations drove the most sales or profit. After the acquisition, Instacart began offering that pricing software to retail clients directly inside its marketplace, effectively turning the grocery app into a live laboratory for Eversight’s algorithms.
Instacart has described Eversight as “an AI-powered pricing and promotions platform,” and it is that engine that is now being unplugged from consumer-facing price tests on the app. The company has told partners that retailers will no longer be able to use Eversight technology to run these kinds of experiments on Instacart orders, a shift that follows scrutiny of how the tool could nudge prices higher by as much as 23 percent for identical items, according to reporting on Instacart and Eversight.
What shoppers actually saw on their screens
For customers, the AI pricing tests did not show up as a flashing banner or an opt-in experiment, they appeared as ordinary prices on everyday items that sometimes did not match what a neighbor or spouse might see for the same product. Some users would simply see the standard price for a given item, while others would be shown a slightly higher or slightly lower price, even when they were ordering from the same store and browsing the same virtual aisle. That meant two people could add the same brand of pasta sauce or laundry detergent to their carts and unknowingly pay different amounts.
The discrepancies came into sharper focus when outside investigators and consumer advocates started comparing receipts. One investigation found that some customers were charged nearly 25 percent more for identical items than others, a gap that raised obvious questions about fairness and transparency in a service that markets itself as a convenient extension of the local supermarket. Another analysis described how the experiment let some users see a slightly lower price while others saw a higher one for the same product, a pattern detailed in coverage of Instacart’s controversial program.
The investigation that forced Instacart’s hand
The turning point came after an outside investigation dug into the real-world impact of the AI-driven pricing tests and found that some Instacart customers were paying nearly 25 percent more for the same items than others. That finding cut through the company’s assurances that the program was about experimentation and optimization, not discrimination, and it gave critics a concrete number to point to when arguing that the tests were effectively penalizing some shoppers for no reason they could see. Once that figure was public, the gap between Instacart’s narrative and customers’ lived experience became much harder to bridge.
In response, Instacart said it would pull the plug on the AI-powered price tests and stop letting retailers run those experiments on its marketplace, even as it continued to defend the broader idea of using data and algorithms to help partners set promotions. The company’s own description of Eversight as an “AI-powered pricing and promotions platform” became a double-edged sword, highlighting both the sophistication of the tools and the risk that they could be used in ways that feel opaque or unfair to shoppers, a tension captured in reporting on how Instacart is pulling the plug.
Customer backlash and the trust problem
Even before Instacart formally ended the tests, the optics of algorithmically tweaked grocery prices were already souring the brand’s relationship with some of its most engaged users. People who rely on the service for weekly staples do not think of themselves as test subjects, and the idea that an AI system might quietly decide they should pay more for the same jar of peanut butter than someone else cuts directly against the promise of convenience and reliability. Once word spread that prices on certain items could vary by as much as 23 percent, the perception hardened that Instacart was letting retailers treat its app as a dynamic pricing playground.
Instacart has tried to counter that narrative by emphasizing that it listens to customer feedback and is willing to change course when experiments undermine trust. Company leaders have said that the decision to end the price tests was driven in part by what they heard from shoppers who were confused or frustrated by the discrepancies, a theme echoed in coverage that described how Instacart ended price tests on groceries after customer feedback and highlighted its acquisition of Eversight in 2022. The company is now leaning heavily on promises of clearer labeling and more straightforward pricing to rebuild confidence.
Wall Street’s reaction and the Eversight question
Investors did not shrug off the announcement. Instacart’s stock dropped after the company said it would stop letting retailers use Eversight technology to run AI pricing tests on its platform, a sign that markets saw the program as more than a minor experiment. For Wall Street, the appeal of Eversight was that it could help retailers squeeze more revenue and margin out of each digital shopper, and turning off that engine inside Instacart’s own marketplace raises questions about how much of the acquisition’s value can still be realized.
The company has not suggested that it will abandon Eversight altogether, and it is likely to keep using the platform in less visible ways, such as helping retailers design promotions or test discounts that apply uniformly to all customers. Still, the immediate stock reaction underscored how central AI-driven optimization has become to the growth story that Instacart has been selling to investors. The tension between that story and the need to keep prices predictable for ordinary families was on display in coverage that noted how Instacart’s shares fell after it stopped AI pricing tests and told retailers they could no longer run those experiments with Eversight technology.
Regulatory heat and the FTC’s $60 million warning shot
Instacart’s retreat from AI price testing is also unfolding against a backdrop of intensifying regulatory scrutiny of its broader business practices. Earlier in December, the Federal Trade Commission announced that Instacart had agreed to pay $60 million in consumer refunds to settle a lawsuit over allegations that it engaged in deceptive tactics around delivery fees and tips. Regulators said Instacart misled consumers by advertising free delivery services and then charging them to have groceries delivered, and that it mishandled tips in ways that shortchanged workers, according to the FTC’s description of how Instacart misled consumers.
That settlement does not directly address AI pricing, but it reinforces a narrative that Instacart has at times pushed the boundaries of what customers and regulators consider fair or transparent. When a company is already writing a $60 million check over allegations of deceptive practices, it has far less room to argue that a program which quietly shows different prices to different users is just harmless experimentation. In that context, ending the AI price tests looks less like a voluntary tweak and more like a necessary step to avoid compounding legal and reputational risk.
Instacart’s promise of clearer pricing labels
In its public messaging, Instacart has tried to draw a bright line between the now-ended AI price tests and its ongoing efforts to make pricing more understandable inside the app. The company has said it will keep clearly labeling when prices on Instacart differ from what customers would see in the physical store, so that people know when they are paying a markup for the convenience of delivery or pickup. That kind of disclosure is meant to reassure shoppers who may have felt blindsided by the idea that the same item could carry different prices for different users.
Instacart has also pushed back on what it calls misconceptions and misinformation about its role in setting prices, emphasizing that retailers, not Instacart, determine the base prices of items and that the company’s tools are meant to help partners experiment within those boundaries. The promise to keep “clearly labeling” price differences is now central to that defense, and it appears in the company’s own explanation of how it is ending item price tests while trying to preserve the parts of its pricing system that customers know and rely on, a stance laid out in its commitment to clearly labeled prices.
What this means for AI in everyday commerce
Instacart’s reversal is a reminder that not every AI application is created equal in the eyes of the public. Consumers have grown comfortable with recommendation engines that suggest recipes or reorder staples, and many accept that surge pricing is part of booking a ride on Uber or Lyft. But when AI starts quietly reshaping the price of a gallon of milk or a box of diapers, the reaction is far more visceral, especially for households that already feel squeezed by inflation and rising service fees. The line between smart optimization and perceived exploitation is thin, and it shifts depending on how essential the product feels.
For other retailers and delivery platforms, the message is clear: algorithmic pricing experiments that touch core necessities will face a higher bar for transparency and consent. If companies want to use AI to test different prices, they may need to move those experiments into more controlled environments, such as loyalty programs where customers explicitly opt in, or limit them to discounts that only move prices downward. Instacart’s experience, including the way its AI pricing tests increased costs for some shoppers even as they lowered prices on certain items, as described in coverage of how Instacart’s AI tests increased costs, will likely be studied closely by any company considering similar tools.
The road ahead for Instacart and its shoppers
Instacart is now trying to thread a difficult needle: it wants to keep pitching itself as a cutting-edge technology company that helps grocers compete in a digital age, while also convincing everyday shoppers that it will not treat their weekly grocery bill as a sandbox for opaque AI experiments. Ending the item price tests is a significant step in that direction, but it will not, on its own, erase the frustration of customers who discovered they had been paying more than others for the same products. The company will need to back up its promises of clearer labeling and more predictable pricing with consistent behavior over time.
For shoppers, the immediate impact is straightforward. The specific AI-driven tests that showed different prices to different users are ending, which should reduce the chances that two people ordering the same item from the same store through Instacart will see unexplained price gaps. Yet the broader questions raised by the episode will linger: how much experimentation are people willing to tolerate in the services they depend on for basic needs, and what kind of transparency do they expect in return? As Instacart moves forward, its success will depend on whether it can convince customers that innovation will serve them, not treat them as data points in an invisible pricing trial.
More from MorningOverview