Search

EvolutionX uses AI Powered Search to power the modern search experience across:

The experience is not the same in every state. Search behavior changes depending on whether:

That distinction is important, because some parts of the experience are cached or personalized, while others are driven by active AI Powered Search queries.

When a logged-out user clicks into the search box without typing, EvolutionX shows a generic search discovery state.

This state can include:

This gives anonymous users a useful starting point before they enter a query.

This experience is primarily cached. It is designed to load quickly and does not depend on a live full-text search at the moment the modal opens.

Data sources include:

AI Powered Search may contribute to the generation of trending products or category associations upstream, but the modal itself is intended to be served quickly from cache.

When a logged-in user clicks into the search box without typing, the search experience becomes personalized.

This state can include:

This helps returning customers get back to items, searches, and categories they already care about.

This experience is built from user-aware data sources, such as:

Some of this content may still be cached, but it is personalized per user.

In this state, AI Powered Search is not necessarily running a live search just because the modal opens. Instead, the UI is usually assembled from user-specific datasets and cached components. AI Powered Search becomes more important once the user starts typing or submits a search.

When the user begins typing, the experience changes from passive discovery to active search assistance.

As the query is entered, the UI can surface:

This helps the customer refine intent before running a full search results page query.

As the user types, AI Powered Search is used to generate:

This is where the advanced search features described below become important.

This is the most search-driven state in the UI and the one most directly improved by AI Powered Search relevance tuning.

The system preprocesses search terms so that customers can still get relevant results even when they type words differently from how products are stored.

If a customer types:

The search layer can generate variations such as:

Customers frequently type shorthand, merged words, or spacing variations. This improves recall without requiring exact formatting.

EvolutionX uses a broader N-gram strategy to improve matching for any search term, not just known compound words.

The search system can derive:

This improves:

The search system can automatically expand queries using synonym dictionaries scoped by store vertical.

A vertical is inferred from catalog context, and the relevant synonym dictionary is applied.

Depending on the vertical, a term like:

Customers do not always use the same language as the catalog. Synonym expansion helps bridge that gap.

EvolutionX search supports term exclusion so users can remove unwanted product types from results.

This is one of the most useful advanced search features for narrowing ambiguous searches.

Dash exclusion

Uppercase NOT

Uppercase & Dash Meaning:

Only uppercase NOT is treated as an exclusion operator.

Lowercase not is treated as a normal search word.

Example

This is treated as a literal phrase-style query, not as an exclusion query.

Bracket expressions allow a customer to exclude multiple terms at once.

Meaning:

Meaning:

This is especially useful when one search term overlaps with several adjacent but unwanted product groups.

The search layer also supports more advanced boolean-style refinement.

Meaning:

This helps when one term has multiple meanings and the user needs to steer the result set more precisely.

EvolutionX search uses AI Powered Search to support both discovery and precision.

The search experience has three distinct states:

The most important search quality features become visible once users start typing, including:

Together, these features make search more tolerant, more configurable, and more aligned with real customer behavior.