From conversational tool to a lifestyle agent: how the Zalando Assistant has grown up
From conversational tool to a lifestyle agent: how the Zalando Assistant has grown up
Three and a half years after its first beta launch, the Zalando Assistant has grown into a companion for the whole shopping journey – from the first spark of an idea to the moment you check where your parcel is. And it now gives you better, more personalised and faster answers than ever. Here’s what’s new, and what we’ve learned along the way.
When we first introduced the Zalando Assistant in spring 2023, it was an experiment with a simple question behind it: what if you could shop for fashion and lifestyle products just by describing what you’re looking for, in your own words? By October 2024, the Assistant was available in all 25 Zalando markets, and with our update in 2025 we made it more personal and better at understanding where you are in the app.
A lot has happened since then. 22 million people visited the Zalando Assistant this year – up from just over 2 million in the first two years. It can also do far more than before. Today, Zalando Assistant can find similar styles from photos and screenshots you upload, check the weekend’s weather and the latest fashion trends, compare products for you, recommend the right size, help you fill your bag and take you to checkout and tell you where your order is, all within the same conversation.
Just as important is the work behind the scenes. Step by step, we’ve improved the quality of the Assistant’s answers and how well it finds the right products, made those answers more personal, and made them arrive faster. The result is simple: you get better, more personalised and faster answers. And customers are noticing – they’re engaging with the Assistant more, liking and adding more items to their bag, and coming back to it more often.
What we’ve learned from our customers
One of the most interesting things we’ve noticed is how differently people behave when they talk to the Zalando Assistant compared with typing into a search bar. Search tends to train us all into a kind of shorthand – “black midi dress”, “white trainers”. In the Assistant, people write the way they’d talk to a friend who knows fashion. Their queries are on average three times longer than traditional search queries, and they ask more than two questions per conversation. Many also pick up on the Assistant’s own ideas: 40% of conversations start with one of its suggestions.
For us, that’s a clear signal. People don’t just want to find a product – they want to talk it through. Much of what we’ve built over the past year comes from that insight.
Weiwei Cheng, Director Agentic Systems at Zalando, says: “When we launched the Assistant, we wanted to find out whether people would enjoy shopping for fashion simply by saying what they’re looking for. Twenty two million users later, we have our answer. Since then, we’ve come a long way in turning it into a trusted advisor that knows your taste and stays with you at every step – and there’s plenty more to come. And with every improvement we make, its answers get better, more personal and faster.”
Show it instead of describing it
Some things are hard to put into words, like the exact shade of a jacket you spotted on someone’s feed. You can now upload a photo or a screenshot into the chat, and the Assistant will find similar items. From there, you can refine the results in your own words – asking for the same look in a longer cut, for example, or in a different colour.
The Assistant has also become more aware of the world outside Zalando. It can now take real-time information into account: the weather forecast for your weekend away, the trends people are talking about right now, or recent events. Planning a weekend hike in the Dolomites? It can check the forecast and tell you whether to pack a jacket – and recommend shoes that are right for the terrain on that particular trail.
And it now goes beyond clothes. We’ve expanded the Assistant’s expertise to beauty, so you can ask it to help you put together a skincare routine or find a makeup shade that matches your skin tone.
Advice that knows your taste – and what you’d rather skip
Personalisation was always our priority, and we’ve taken it a step further. Since mid-July, the Assistant uses what it has learned about your preferences – like the brands, colours and materials you tend to go for – to shape not only which products it shows you, but also how it talks to you and explains its suggestions.
Knowing what you don’t want is just as useful. Just say what you don’t want – a jumper without synthetic materials, a t-shirt but not in black, or results that skip a particular style altogether – and the Assistant filters it out. And once you’ve narrowed it down to a few options and can’t decide, you can compare products side by side, right in the chat.
Found your style? Let’s find your fit
Finding the right product is one thing. Finding the right size is another. That's why we're bringing Zalando's size advice directly into the Assistant. Once you've found something you like, you can now ask which size is most likely to fit you – without leaving the conversation.
It might sound simple, but there's a lot of technology behind that answer. General AI assistants can suggest what to wear or where to buy, but knowing which size will actually fit you is a different challenge. Since 2018, we've built technology that maps more than 10,000 size charts across hundreds of sizing systems, making sense of how sizes differ between brands and products.
By combining this with insights from millions of customer orders, what they've kept or returned, and even their own body measurements, we can provide personalised size recommendations that go far beyond a generic size chart.
Today, our Size & Fit solutions cover around 70% of Zalando's assortment. And by bringing this intelligence into the Assistant, we're making the journey from finding something you love to finding the right fit that much simpler. By prioritising products available in customers’ sizes, the Assistant has already increased add-to-cart rates by 4% in A/B testing.
From conversation to checkout – and beyond
Until recently, the Assistant would help you find something and then hand you over to the rest of the app to buy it. Now you can get your bag ready in the conversation itself: add or remove items from your bag, check sizes and availability, and go to checkout without ever leaving the chat.
The help doesn’t stop once you’ve placed your order, either. Ask the Assistant where your order is, and it answers with an order card showing exactly where your parcel is. People clearly prefer it: since we rolled the cards out in July, 92% of customers checking on an order go on to interact with the card, compared with 65% who engaged with the old text-only replies before.
Where it goes from here
Much of what makes the Assistant useful comes from our own technology. Zalando’s proprietary AI models handle product search, recommendations, visual discovery and size advice, built on 18 years of experience in fashion e-commerce, behavioural and Size & Fit insight. We combine them with large language models from OpenAI and other leading AI models to power the conversation.
Graeme Smith, SVP, Chief Technology Officer at Zalando, says: “AI is changing the way people discover fashion, and we want Zalando to be where that journey starts. The Zalando Assistant plays an important part in our bigger vision of Lifestyle AI. Think of it as a layer of intelligence that runs through the whole Zalando experience: over time, it gets to know you – your taste, your size, the occasions you’re shopping for and what you need right now. Whether you’re chatting with the Assistant, scrolling your Feed or searching for something specific, every interaction helps us get to know you better than ever before. We use that to create an experience that feels like a true fashion and lifestyle companion – one that knows you and the world of fashion and lifestyle inside out.”
Key facts at a glance
- 22 million people visited the Zalando Assistant in 2026
- Better, more personalised and faster answers: continuous improvements to answer quality, product search, personalisation and speed, leading to higher engagement, more likes and add-to-bags, and more returning users
- Available in all 29 Zalando markets, in local languages, on app and web
- Queries are 3x longer than in traditional search, with more than two questions per conversation on average
- 40% of conversations start with a suggested prompt
- Built on our own architecture using OpenAI’s large language models, and powered by 18 years of experience in European fashion e-commerce, behaviour and Size & Fit insight