AI search: how to improve product search
This article will explain the principle of AI search in Fynode, specifically how the module understands the context of queries, handles synonyms, and why it gives better results than classic full-text search. We will describe integration via JS code, settings to improve relevance, and practical steps to test and measure the impact on conversions.
How AI search works
The AI module uses models that understand the meaning of words and the context of queries, enabling them to find the right products even with typos or synonyms. The system takes into account product popularity and sales to prioritize the most relevant results for the customer. Integration and management of the plugin is done via the AI plugins section in the administration.
Improving relevance: synonyms and filters
Enable synonym generation or use a manual synonym database to cover common variant terms (e.g., mixer = whisk). Activate the filter “Show only products that are in stock” if you want to prioritize available items. Test results on real queries and fine-tune word mapping and relevance weights based on customer behavior.
Integration into the eshop (JS code)
After activating the plugin you will get generated JS code to insert into your eshop (head or footer depending on the platform). This allows the AI search to run directly on the site without further intervention. Different platforms support code insertion differently, so follow the instructions for your platform in the eshop connector documentation.
Metrics, budget and deployment
Use A/B testing, measure CTR from search results and the conversion rate from search. The plugin settings include a calculator to estimate costs based on the number of searches, so plan the budget and call frequency. Before deploying to the entire catalog test the plugin on a small sample of products and monitor AI credit consumption.
After inserting the JS code and activating the plugin, search is immediately available in the eshop. The plugin may perform additional indexing or synonym generation which can take several minutes, so we recommend first checking the preview and then running tests on a sample of users.
In the plugin settings use the section for generating or managing synonyms. The system can run through products once and create a synonym dictionary that will improve search matches. Synonym generation is priced and you will see an estimated credit consumption before starting the process.