AI-driven discovery risks becoming a returns engine, MACH Alliance warns

By Tom Whitfield

Retailers using AI to lift front-end conversions could be manufacturing a second problem downstream: 45% of shoppers say unrealistic expectations created by AI discovery would make them more likely to return a product. The figure comes from a poll of 1,000 consumers by the MACH Alliance, the vendor-neutral body that sets standards for open, composable and connected enterprise architecture.

The commercial logic is straightforward. AI already shapes a large share of digital demand. Some 42% of shoppers' digital spend is influenced by AI search, rising to 48% among Millennials. Agents directly influence 41% of ecommerce purchases, and 45% of social commerce spend. Discovery, in short, is increasingly mediated by machines.

The catch sits at the doorstep, not the checkout.

What does the returns risk actually look like?

The numbers are consistent across cohorts:

• 56% of shoppers think AI overhyped their product expectations before checkout.
• Six in ten say AI creates unrealistic beliefs about how products will look, fit or perform; 44% say it sets expectations too high, rising to 47% of Gen Z.
• 68% agree that "too perfect" AI imagery often fails to reflect the product that arrives.
• 60% think AI-generated styling or visualisations make it harder to imagine products in real life.
• 51% say using AI increases the likelihood of buyer's remorse post-purchase.
• 53% of Gen Z say AI makes them more likely to return products.

For sourcing and ecommerce teams, that reads as a reverse-logistics cost curve bending upward in proportion to AI-driven conversion gains. Every incremental order won through over-polished visuals carries a probability of coming back.

Whose problem is it to fix?

Jason Cottrell, president of the MACH Alliance, was blunt about the mechanism. "When the product that arrives doesn't live up to what AI promised, that sale quickly becomes a return," he said.

His prescription is architectural rather than cosmetic. "Agents need to be grounded in accurate, real-time product information and connected to the wider commerce ecosystem, from inventory and availability to fulfilment and post-purchase. No single vendor or agent can solve for every part of that journey."

In practice, that means product data quality, catalogue accuracy and integration depth move from back-office concerns to returns-prevention levers. An agent recommending a garment against inventory, sizing and imagery data that reflects the shipped product is a different proposition from one generating aspirational visuals detached from the SKU.

Cottrell again: "Connected ecosystems are how retailers close the AI expectations gap. When agents, data and best-of-breed technologies work together, what shoppers see during discovery is much closer to what arrives at their door."

Confirmed results or marketing framing?

Two caveats are worth noting. The sample is 1,000 consumers — directional, not exhaustive. And the research comes from a body whose stated purpose is promoting composable, connected architecture; the finding that disconnected AI causes returns conveniently argues for its members' approach. The data on expectation gaps is self-reported consumer sentiment, not measured return rates.

What is confirmed is the influence side of the ledger: AI is already steering between 41% and 48% of spend depending on channel and cohort, and returns exposure scales with it.

What does this change for sourcing and supply-chain professionals? It adds a data-integration mandate to the AI roadmap — accurate, real-time product information feeding every agent touchpoint — and it puts the cost of polish, in pounds per return, on the agenda of whoever owns the AI budget. Brands that treat AI visuals as a marketing layer rather than a product-data problem will likely learn the difference through their returns departments.

Expect the expectations gap to feature in 2025 planning cycles, whether or not anyone names it that.