A new study from Consumer Reports (CR) alleges that Instacart is conducting AI-driven pricing experiments that inflate product costs for some customers. The research indicates that shoppers at major retail partners like Kroger, Albertsons, Costco, and Safeway may be paying as much as 23% more than others for identical items.
The AI Pricing Tool
The experiments are powered by Eversight, a software-as-a-service (SaaS) product designed to help grocers maximize revenue through strategic pricing. Instacart admits that some shoppers will see higher prices, but the Consumer Reports study suggests the increases are significant—not merely “slight,” as the company implies.
How It Works
The pricing adjustments are conducted through limited online tests, involving 10 U.S. retail partners that already apply markups. Instacart defends the practice by claiming it mirrors traditional in-store price testing. However, the AI-enabled system raises questions about transparency and fairness in digital marketplaces.
Instacart’s Response
The company insists these are not real-time “dynamic pricing” adjustments, but rather randomized experiments unrelated to individual shopper behavior or market demand. Instacart emphasizes that prices remain fixed during the tests. Still, a 23% price increase is substantial and raises concerns about potential exploitation.
The Broader Trend of Dynamic Pricing
Instacart’s practice is part of a larger trend of AI-driven dynamic pricing across e-commerce. Other major retailers, including Amazon, have been accused of similar tactics. A recent report alleged that Amazon’s dynamic pricing inflated school supply costs for districts nationwide. These practices are becoming increasingly common, but often lack clear oversight.
The Consumer Reports study highlights a growing issue in online retail: AI-powered pricing experiments that could disproportionately impact consumers without their knowledge. The lack of transparency surrounding these tests raises questions about fairness and the potential for hidden markups.
