Text Search

Text Search powers your shop’s search bar. It combines semantic (meaning-based) matching with exact keyword matching, so a shopper typing “red winter jacket under 50€” gets relevant results even if no product title contains that exact phrase, while a shopper pasting in a SKU or product name still gets an exact hit.

How it works

flowchart TD
    A[Shopper types a query] --> B[Query understanding\nbrand, price, size, color]
    B --> C{Search}
    C --> D[Semantic match]
    C --> E[Exact match\nID / GTIN / title]
    D --> F[Merge & rerank]
    E --> F
    F --> G[Filters, boosting,\ndiversification]
    G --> H[Results shown to shopper]

What it does

🧠 Understands the query

Free-text queries are analyzed to detect brand, category, price, color, and size — “red winter jacket under 50€” becomes color=red, a price ceiling, and a jacket-type category. Recognized signals are applied automatically as filters or offered as a suggested filter chip.
🎯 Combines semantic and exact matching

Products are ranked by meaning-based similarity to the query; products that exactly match an ID, GTIN/barcode number, MPN, or title are matched directly and can be shown ahead of (or instead of) the semantic results.
🔀 Smart reranking

Semantic and keyword-match signals are blended so results that closely match the literal words of the query aren’t buried under purely “similar meaning” results.
🌈 Result diversification

For broad queries, results can be spread across categories instead of showing many near-identical products.
⌨️ Auto-suggest

As a shopper types, the search bar can show live query completions drawn from your own catalog.
↪️ Redirects

A specific query can be mapped to redirect straight to a URL — useful for well-known searches you want to send to a dedicated landing page.

Configuring in Panel

In Panel, open Site → Campaigns, and either open an existing Text Search campaign or create a new one (Campaign Type: Text Search). A new campaign lets you start from a default configuration, duplicate an existing campaign, or configure everything manually.

General settings — Name, an Active toggle, and a Default Campaign toggle (the campaign used when a request doesn’t specify one; only one campaign per type can be default).

Methods

Semantic Search — turn on the AI models (“Nets”) this campaign should use and set each one’s weight (e.g. Berry Punch Net, Volcanic Water Net, Darjeeling Second Flush Net). A Net can only be weighted here once it’s turned on for the site under AI methods. Inside this section, Minimum Similarity Difference and Maximum Similarity Difference exclude results that are too close to (likely duplicates) or too far from the query; Minimum Pairwise Difference keeps returned results from looking too similar to each other.
Auto-Suggest — toggle to enable query completions for this campaign’s catalog.
Query-to-URL Mapping — add specific query strings, each paired with a URL to redirect to.
Query-to-Filter Automation — toggle to let recognized query signals (brand, size, etc.) auto-apply as filters rather than only being suggested.
Smart Reranking — toggle to blend keyword-match strength into the ranking, on top of semantic similarity.

Influencers

Exact Search — turn on Exact Match First to match IDs, GTINs, and MPNs directly ahead of semantic search; within it, Exact Search Only restricts results to exact matches alone.
Result Diversification — spread results across categories for broad queries.
Only Buyable Products — restrict results to in-stock products.

Interface controls

Preferred Image — which product image to show in results (first or best_match).
maximum results / maximum exact results — caps on returned results.
Displayed Dynamic Filters — how many filter panels to show alongside results.
Filters — opens a Filters panel with an edit tab (turn on which catalog attributes are filterable, give each a custom label, and mark some as Primary) and a sort tab (order filters by relevance or set a manual drag-and-drop order).

Products — restrict the campaign to specific categories via Assigned Products, or leave it unrestricted to search your whole catalog.

Boosting — link one or more Boosting campaigns and set how strongly they influence ranking. See the Boosting page for the full configuration.

If item grouping is on for your site, variants of the same product (e.g. different sizes) are shown as one result, with the other variants available alongside it.

MCP Tool

Agents can call this service directly as the text_search MCP tool, which wraps the Text-Search Products REST operation — its request and response shapes match that operation’s schema.