Customers do not always search using the exact terms you use in your catalog. For example, they might search for "outfit for a wedding" while your product is named "cocktail dress".
With AI Search, the Tweakwise search engine understands what the visitor means, rather than solely matching literal keywords. As a result, visitors find the right product faster, encounter fewer zero-result searches, and conversion rates across your webshop increase.
In this article, you will learn how to configure AI Search, what you need to get started, and how Tweakwise continuously improves behind the scenes.
1. Getting Started: Setting up AI Search
Adding AI Search to your setup is straightforward via your existing search configuration:
In the Tweakwise App, navigate to Search > Search algorithms.
Add the AI Search module to your search algorithm.
Configure the following settings:
Enabled: Determine whether the algorithm is active immediately.
Tip: Leave this unchecked initially if you prefer to test the results safely in the Demo Shop first.
Accuracy: Use this slider to determine how strictly the model filters results:
Low accuracy: Returns more results, but with a broader interpretation (may include looser suggestions).
High accuracy: Returns fewer results, but the items displayed match the search query with high precision.
(When added, Tweakwise automatically sets the accuracy to our recommended default score).
Stop / Continue: Determine what happens once AI Search finds matching products:
Stop: Display these results immediately and do not evaluate subsequent algorithms.
Continue: Combine the AI Search results with the next algorithms in your list.
Weight: Use the slider on the right to set the relative weight of the AI Search algorithm compared to your other rules.
2. What do you need to get started?
To enable AI Search effectively for your specific product catalog, two components are essential:
A. Tracking events
The AI model learns directly from real search and click behavior on your webshop. The following events must be tracked via the Tweakwise Event Tag:
Searches
Result views (PageImpressions)
Product clicks or visits (ItemClicks / PageViews)
B. Product data
The richer and more complete your product attributes are (such as color, fit, material, or category), the better the model understands which items match a query.
Tweakwise automatically combines your primary categories and product attributes into an optimized description (e.g., "Adidas Ultraboost running shoe in category Men's Shoes with color Black and cushioning Responsive").
3. How does it work behind the scenes?
You do not need to build complex machine learning models yourself; Tweakwise handles the heavy lifting in the background:
Learning from real behavior: When a visitor searches for "light summer shoe" and clicks on a specific pair of canvas sneakers, the system records this connection. As this pattern recurs, the model learns that this sneaker is a strong match for that search query.
Smart from day one, even with low data: If your webshop currently has low search volume or you are launching an entirely new category, our AI automatically maps logical search queries and synonyms to your product descriptions. This ensures AI Search provides relevant results from day one.
Smarter every week: The AI search model is retrained automatically on a weekly basis using the latest click and interaction data. This enables the algorithm to adapt seamlessly whenever your product catalog updates or customer search behavior shifts.