Zalando’s AI Fashion Assistant: A Better Question Than “Show Me More”

By Kirill Naumov · 2026-10-25

Fashion shoppers rarely arrive with a perfect product name. They have a wedding next month, a suitcase with limited space or a coat that needs to work with the clothes they already own. A catalogue knows categories. A good adviser understands the occasion—and then checks whether the recommendation can actually be bought.

Zalando said its AI-powered assistant became available to logged-in customers across 25 markets in local languages in October 2024. Its March 2025 update connected the assistant more closely to account preferences and shopping history, and made it aware of where a customer was browsing. These are Zalando’s public product announcements, not a project by Kirill Naumov or a claim that every recommendation was accurate.

Context can save a question—or make a wrong assumption

If a shopper opens the assistant while browsing denim, suggesting a question about jeans is more useful than beginning with a blank chat box. But page context is a clue, not an instruction. The person may be looking for a gift, researching sizes for someone else or simply comparing styles. An assistant should make it easy to correct the premise rather than insist that the last click defines the customer.

Shopping history raises the same issue. A previous purchase can suggest a preferred size or brand, but it may have been returned, bought for a relative or made years ago. Personalisation earns trust when the reason for a suggestion is understandable and the customer can change it. “Because you viewed this” is not a personality profile.

Make the product data do the hard work

A conversational answer can describe an outfit beautifully and still fail at the point of decision. Does the recommended item exist in the customer’s size? Is its colour represented accurately? Will it arrive before the event? What is the return condition? Those answers come from product, stock, delivery and policy systems—not from a model’s confidence.

For a pilot, choose one shopping mission, such as finding an outfit for an occasion. Check the assistant’s suggestions against approved product attributes and live availability. Compare the journey with ordinary search and filters, including the time a customer spends correcting unsuitable recommendations. The goal is not a longer conversation. It is a shorter path to a confident choice.

Read engagement numbers carefully

Zalando reported that more than two million customers had used the assistant by March 2025. It also reported a 40% increase in “high-value interactions” during a pilot of the updated experience, with examples including likes and add-to-cart actions. Those are company-reported engagement measures. They are not, by themselves, evidence of incremental sales, fewer returns or a better outcome for every customer.

A retailer evaluating its own assistant should also look at completed purchases, unsuitable-item returns, unanswered questions, customer corrections and whether colleagues can resolve a failure quickly. Compare like-for-like customers or sessions and document changes in traffic, promotions and assortment. If a simpler product guide performs just as well, the chatbot has not yet earned its upkeep.

The useful lesson from Zalando is not that every fashion retailer needs a chat interface. It is that the journey can start with a person’s situation rather than a rigid taxonomy. The conversation matters only when the business can support it with trustworthy choices, honest uncertainty and a clean route to purchase.

Explore the English AI for Retail workshop → · Discuss a retail AI consulting brief

Primary sources: Zalando’s October 2024 rollout and March 2025 update. Figures are company-reported, not independent proof of incremental sales.

Start a conversation →