How it works
Your WordPress search still works. We add to it, never replace it.
Every exact match and keyword hit your site already finds appears first, in the same order. Semantic matches are added below — results your shoppers would describe but your product titles never literally contain.
See it running on 3 sample stores →Type your own query. Compare default WordPress search vs. AI Search side by side.
Same query. Default search misses three sales.
A real example from a clothing catalogue. Default WordPress search returns only products whose titles literally contain the words. AI Search returns those too — and three more that match in meaning.
Query: “cozy winter sweater”
Merino Wool Sweater — Classic Crew
Cashmere Sweater — Slim Fit
2 results · keyword matches only
Query: “cozy winter sweater”
Merino Wool Sweater — Classic Crew
Cashmere Sweater — Slim Fit
+ Chunky Knit Pullover — Oversized
+ Cable-Knit Cardigan — Wool Blend
+ Lambswool Jumper — Camel
5 results · same shop, three more sales
Three tiers, in strict order.
Every search runs through all three in sequence. A tier may add results; it never replaces results from the tier above. The merge is deterministic — same query, same site, same ranking.
TIER 00
Exact
SKU matches and exact product titles. Pinned at the top with a 1.0 score, no matter what.
TIER 01
Keyword
Everything WordPress and WooCommerce search would find natively — the words literally appear in the title, content, or taxonomy. Re-ranked by relevance, not date.
TIER 02
Semantic (AI)
Products that mean the same thing even when they use different words. Added below the keyword set, never swapped in for them.
Why this order matters. If a shopper searches for an exact product name, they get that product first — the AI never second-guesses an obvious match. Keyword matches follow, then semantic matches extend the list. Installing AI Search does not change the results your customers find today. It only adds the ones they would have found if your product titles happened to use their exact words.
Smart Matching · Standard plan
The invisible rewrite that makes semantic search accurate.
Before each product is indexed, our LLM reads it and writes 3–5 alternative phrasings a shopper might use. That expanded version is what the vector index sees — so queries like cheap sofa start matching products titled affordable couch. Up to 60% better semantic results on everyday shopper queries.
What you saved
Hoodie Oversized — Brown
What the index sees
Enrichment runs once per item, when you publish or sync. The result is cached forever — your shopper’s search is never slower because of it. See the Standard plan →
Zero latency added at search time.
Everything expensive — reading your products, running the LLM, generating vectors — happens asynchronously, when a product is published or re-synced. By the time a shopper searches, the index is already warm.
Index time · once per item
LLM enrichment
Smart Matching
Vector generation
Per product
Index store
Post meta + cache
Search time · every query
Query embedding
3 s timeout, falls back
Cosine similarity
Up to 2,000 products
Tier merge + render
Deterministic
See it live before you install.
Three WooCommerce demos: clothing, bookstore, electronics. Side-by-side comparison with default WordPress search. No sign-up, no plugin install — your own queries, instant results.
Try the live demo →