Structured and exploratory e-commerce search: Keeping the AI from hallucinating

In the fast-paced world of e-commerce, the success of your business hinges on one simple question: Can your customers find what they’re looking for? If not, they’ll be off to a competitor faster than you can say “abandoned cart.” 

The emergence of AI powered search engines are changing the game rapidly. But the practical and effective use of AI search is tricky as it needs to do two things simultaneously – reap the benefits of AI while keeping the sanity of your normal, traditional search. 

Two critical types of search behavior – structured and exploratory – play a significant role in ensuring customers find what they need, in the era of AI driven search. Mastering both can dramatically enhance your product findability, and ultimately, business performance. 

Structured Search: Find precisely THAT item

Structured search is like a customer who knows exactly what they want. They enter specific criteria, like a brand name or product model, and expect precise results. Imagine someone searching for “Nike Air Max 270” – they’re not looking for just any sneakers; they want that sneaker.

Structured searches are all about precision. A solely AI based search solution is prone to hallucinate, and find items that just “look like” Air Max 270. This can be frustrating to the shoppers, and we don’t want that. 

Stay in the Know - Subscribe for AI Innovation and E-commerce Trends

Join the Miros community and receive expert advice on driving growth through AI-powered product discovery.

Exploratory Search: Find anything you have in mind

Now, exploratory search is a whole different ballgame. Think of it as window shopping in the digital age. Customers might not know exactly what they want, but they’re browsing broader categories, like “summer dresses”, “Top Gun sunglasses” or even “cool gadgets I didn’t know I needed.” No product is labeled as such, but the AI search can certainly find them. 

In this scenario, the AI search engine needs to play the role of a savvy salesperson—presenting a variety of relevant options to catch the customer’s eye. It’s about turning that casual browser into a buyer by encouraging product discovery. The more options you show, the higher the chances of them thinking, “Well, this is better than I expected to find.”

Screenshot: Miros combined exploratory/structured search solution in action. No lamps have been labeled as “industrial” – it’s all AI. 

The Magic of Combining Lexical and Semantic Search

To nail both structured and exploratory searches, a great search solution has to blend lexical and semantic search capabilities. 

Lexical search is straightforward – it matches the exact words the customer types into the search bar with your product descriptions. Perfect for those structured searches when someone knows exactly what they want.

Semantic search, on the other hand, is like the mind reader of search engines. It understands the intent behind the search, even when the customer’s language is vague. When someone searches for “beachwear,” semantic search for e-commerce knows they’re not just looking for a swimsuit but might also want towels, hats, and flip-flops. It’s like knowing what your customer wants before they do.

And to make things harder, many searches actually combine structure and exploration. Think of “vintage Ray Ban sunglasses”. Many vintage Ray Bans are not labeled “vintage”, however the AI recognises them very well as “vintage” – so this is your exploratory search. On the other hand, “Ray Ban” is a discrete brand match, so the results should show Ray Bans only. This is structured.  

The Technical Challenge (Spoiler: It’s Not Easy)

Creating a system that handles both structured and exploratory searches isn’t a walk in the park. It’s more like a tightrope walk, balancing precision with flexibility. What needs to happen? The entire search query first needs to be analyzed by one AI, to understand which words or phrases are of structured, and which of exploratory nature. Then multiple queries need to happen on the back end, to fetch both lexical and semantic search results – fast. And then all these sources of information should be combined to one, attractive search results page. All in a few hundred milliseconds. 

Driving Revenue with Better Search Experiences

Optimizing both types of searches has a direct impact on your bottom line. Structured searches drive up conversion rates by delivering precise results, while exploratory searches boost AI-driven product discovery, leading to more items in the cart and higher average order values.

What now? Talk to us to see how Miros can solve this for your business. 

Stay in the Know - Subscribe for AI Innovation and E-commerce Trends

Join the Miros community and receive expert advice on driving growth through AI-powered product discovery.

Success! Your request has been submitted!​

We’re thrilled that you’re interested in experiencing how Miros can enhance your eCommerce strategy. Our team is already reviewing your request, and we’ll be in touch soon to discuss the next steps and schedule your personalized demo.