When the AI Becomes the Store Associate: Why a 375% Spending Increase at Macy's Is Both Exciting and Alarming

Macy's launched "Ask Macy's" across all its digital platforms in late March 2026. The tool is powered by Google's Gemini and works as a conversational shopping assistant — describe what you need, and the AI recommends products, suggests complete outfits, and offers virtual try-on. Bloomberg reported the testing data: roughly half of Macy's website visitors were exposed to the chatbot over several weeks, and those who engaged with it spent approximately 4.75 times more than those who did not.

I have been building Mubboo as a comparison platform across five countries. The Macy's result is the most concrete evidence yet that AI shopping assistants change consumer behavior at scale. A 375 percent spending increase is exciting because it suggests AI can genuinely help consumers discover products they want. It is alarming because the AI doing the recommending was built by the company selling the products. Those two facts need to be held together, not separated.

The department store experience is being rebuilt in software. The question is whose interests the new version serves.

What Makes Ask Macy's Different from a Search Bar

Traditional online shopping works like this: customer types keywords into a search bar, gets a grid of products, scrolls through dozens or hundreds of results, and decides. The cognitive load is on the consumer. The retailer provides the inventory; the shopper does the work.

Ask Macy's inverts that dynamic. A customer describes an occasion, a need, or a style preference in natural language — "I need shoes for a special occasion" or "Help me find a gift for my mother." The AI interprets the request, recommends specific products, and then extends the conversation. The "Complete the Look" feature is the key driver of the spending increase. A customer shopping for a navy blazer gets prompted with a matching shirt, then a belt, then shoes, then a pocket square. Each suggestion builds on the last. The interaction is conversational, personal, and directive in a way that a product grid never is.

PwC's Ali Perhman described the distinction: "Effective AI is closer to a personal shopping agent that understands product composition and customer tastes, rather than just a chatbot." Macy's invested accordingly. Thousands of employees contributed to refining the system before public launch. Early versions produced mechanical, list-style responses. Updated versions use conversational prompts — "Do you prefer bright colors or calm colors?" — and account for regional weather differences, brand surfacing priorities, and seasonal relevance. This is not a generic chatbot bolted onto a website. It is a merchandising tool with a conversational interface.

AI as Discovery Engine

Online shopping has always suffered from a discovery problem. A typical department store website lists tens of thousands of products across hundreds of categories. For consumers who know exactly what they want, search works. For consumers who have a vague need — "something for a wedding next month" — the experience is overwhelming. Too many options, too many tabs, too much cognitive effort just to narrow the field.

A well-designed AI shopping assistant addresses this directly. Instead of browsing 200 dress options, the consumer describes what they need and gets a curated selection of five or six. The AI acts as a filter, translating vague intent into specific, purchasable recommendations. For consumers who struggle with decision fatigue — which is most consumers, most of the time — this is genuine value.

The 4.75x spending increase may partly reflect consumers finding products they actually wanted but would not have discovered through traditional browsing. A customer who came to buy a single item but left with a coordinated outfit may be a customer who got a better result, not a customer who got manipulated. That distinction matters.

The broader industry is moving in the same direction. PwC data indicates approximately 40 percent of the top 20 US retailers by revenue have now deployed some form of AI shopping assistant, with most launches concentrated from mid-2025 onward. Macy's is not an outlier — it is an early mover in what is becoming the default retail interface. Store associates at physical Macy's locations have begun using similar AI capabilities to assist in-person customers, extending the technology beyond digital channels.

Consumers interact with AI shopping assistants on their phones without thinking about who built the AI or what it was optimized for.

Who Does the AI Serve?

Here is the tension at the center of this story. The AI that advises the consumer what to buy was built, trained, and optimized by the retailer that sells the products. Ask Macy's is not an independent advisor. It is a sales tool with a conversational interface.

"Complete the Look" recommendations increase basket size by design. The AI suggests complementary products because Macy's wants larger orders — that is the explicit business objective. The consumer did not ask for cross-selling. They asked for a blazer. The shoes, belt, and pocket square are the AI's addition, surfaced because they increase average order value.

The 4.75x spending increase is measured against non-AI users. But the question consumers should ask is: did I spend 4.75 times more on things I genuinely needed, or 4.75 times more on things the AI persuaded me to add? The answer for any individual customer depends on the specific interaction. At population scale, the answer is almost certainly both — some genuine discovery, some optimization-driven upselling, blended together in a way that is difficult to separate.

This is not a hypothetical concern. Every recommendation the AI makes is informed by Macy's inventory, margin targets, and brand partnerships. The model was refined to "surface top brands in dress recommendations" — an editorial choice that serves the retailer's commercial relationships, not the consumer's budget. The system was trained by Macy's employees to reflect Macy's merchandising priorities.

At Mubboo, we recommend products based on editorial judgment and consumer research, not retailer margin. We include anti-recommendations — "skip this" — when a product does not justify its price or when a better alternative exists. A retailer's AI assistant will never tell you not to buy something. That is the structural difference between a platform built to sell and a platform built to compare.

What This Means for Comparison Platforms

Retailers building their own AI advisors is a defensive strategy. They want to keep consumers on their sites instead of losing them to ChatGPT, Perplexity, or Google's AI Mode for product discovery. If a consumer gets their shopping advice from a platform that is not the retailer, the retailer loses control of the recommendation — and potentially the sale.

This creates a new dynamic for consumers. Every major retailer is becoming its own AI-powered recommendation engine, each optimized for that retailer's inventory and commercial objectives. A customer asking Macy's AI for a blazer recommendation and Nordstrom's AI for the same request will get different answers — not because the products differ, but because the optimization targets differ. Macy's AI surfaces Macy's brands. Nordstrom's AI surfaces Nordstrom's brands. Neither tells you that a better option exists at the other store, or at a third retailer, or at a fraction of the price from a direct-to-consumer brand.

At Mubboo, our Shopping channel across the US and Australian sites exists to provide that independent layer. We compare products across retailers, include honest assessments of quality relative to price, and structure our content for both human readers and AI agents that need trustworthy data sources. Our editorial judgment is not shaped by any retailer's inventory or margin targets — and that independence becomes more valuable as every retailer deploys its own AI advisor.

The Macy's result is a preview of where retail is headed. Every major retailer with an AI assistant. Every assistant optimized for that retailer's business objectives. Consumers navigating a landscape where the "advice" they receive is commercially motivated by default. Independent comparison is not competing with that system. It is the check on it.

The Store Associate Analogy Is More Accurate Than Macy's Intended

Macy's told customers to interact with the AI "like you would a store colleague." That framing is more revealing than they may have intended. A store associate works for the store. Their incentive is to increase basket size, move inventory, and build brand loyalty. A good associate genuinely helps — suggesting items that work, steering you away from poor fits. But even the best associate operates within the store's commercial interests.

The AI shopping assistant is the same, at scale. It helps millions of customers simultaneously, and many of them will get genuine value. But it works for Macy's. The recommendations serve the store first and the shopper second. That has always been true of retail — the AI did not create the dynamic, it amplified it.

What is new is the interface. A human store associate is visibly employed by the retailer. The commercial relationship is obvious. An AI chatbot that says "tell me what you are looking for" feels like a personal advisor, even when its recommendations are shaped by margin targets and brand partnerships. The conversational format obscures the commercial motivation in a way that a product grid never did.

For consumers, the response is not to avoid AI shopping assistants — they are genuinely useful. The response is to check the recommendation against an independent source. That is what we are building at Mubboo, and the Macy's result confirms that the need for it is growing, not shrinking.

Sources: Bloomberg (March 26, 2026), Fortune (March 27, 2026), PYMNTS (March 26, 2026), PwC, BigGo Finance.