In February of this year, Shopware announced the integration of an AI-based search in collaboration with Nosto . This is particularly relevant for the German market, where Shopware is the leading e-commerce platform (EHI). It is also time for us to take a broader look at the topic in general. For over 15 years, Comsysto Reply has been implementing complex e-commerce solutions for clients across all industries. In parallel, we track and analyze relevant trends in our e-commerce lab to ensure we are ready today for the challenges of tomorrow. We share our insights here—every Tuesday!
You can learn more about the collaboration between Shopware and Nosto in the E-Commerce Magazin or directly at Shopware . Other e-commerce platforms like Shopify are also integrating semantic product search into their software.
What exactly is special about an AI search solution or semantic search?
The main difference is that the customer can interact with the system using "natural language," which is often referred to as NLP (Bloomreach) is what it’s called. However, how “natural” the language can actually be depends on the maturity of the system in question. Fundamentally, it doesn’t just search for keywords; it also considers the context in which those words are used. This allows for an analysis of semantics—the actual meaning of the words. For example, “strapless” and “off-the-shoulder” can be treated as synonyms, at least when it comes to dresses. Furthermore, more sophisticated systems can interpret entire sentences or even incorporate recommendations.
This entire field is not new; it has been around for years. Currently, however, the topic of NLP is gaining significant momentum because LLMs (Large Language Models) are exceptionally well-suited for understanding human language. And there have been enormous advances in this area over the last few years.
So, how does it work in practice?
Our engineering team at the Comsysto Reply e-Commerce Lab has also been exploring how to implement such solutions. In our case, we used OpenAI, the provider of ChatGPT, to first derive relevant classifications from a human query. What is remarkable about this example is that the AI can correctly derive classifications like “Season: Summer” or “Length: Short” from a purely textual description. The latter, in particular, is not explicitly mentioned in any way; the criterion is derived purely semantically.
These classifications are then linked using logical operators (AND, OR, NOT), and the resulting data structure is transformed into a search query for the e-commerce platform being used. A more detailed description of the concept and the technical implementation can be found in our blog .
Of course, this is only the first step. Other important aspects to consider subsequently include sorting search results through appropriate scoring or ranking, and interspersing recommendations that are closely related to the items found.
That’s it for today—and here’s what’s next
In general, we are moving toward a future where traditional keyword-driven searches and category pages with filters are increasingly being replaced by more flexible and customer-friendly operating concepts. And it is certainly relevant not just to use these, but to at least roughly understand how they work—at the very least for online shop operators who have to choose one solution or another.
What won’t change so quickly: once the right products are in the shopping cart and ordered, they still have to get to the customer somehow. This remains one of the most complex problems in e-commerce. And that is exactly what we will be looking at in the next edition.
See you next week!
Christian & the Comsysto Reply e-Commerce Lab