AI-Powered Search

Your Customers Don't Search in Keywords —
Why Should Your Search?

Ultrafast Search combines Meilisearch keyword matching with OpenAI semantic embeddings. Your customers search naturally — the AI understands what they mean, not just what they type.

Start Free Trial →
OpenAI Powered Built into every plan Full cost control
AI expanding query...
The problem

Keyword Search Fails When Customers Think in Sentences

Traditional search only matches exact words. When a customer types “comfortable shoes for standing all day,” keyword search returns nothing — because no product name contains that phrase. Every failed search is a lost sale. Customers leave; they do not try a second query.

Zero results, zero sales — 15% of e-commerce searches return nothing, and each one drives a customer away
Synonym blindness — keyword search cannot connect “footwear” to “shoes” or “couch” to “sofa”
No intent understanding — “something warm for winter hiking” is meaningless to a keyword engine
Misspelling limits — typo tolerance fixes 1–2 characters, but it cannot bridge vocabulary gaps
Search Comparison
Traditional keyword search
comfortable shoes for standing all day
0 results found
Ultrafast Search with AI
comfortable shoes for standing all day
CloudWalk Comfort Sneaker €89.99
StandEasy Pro Work Shoe €119.00
UltraCush Daily Trainer €74.50
ArchSupport Classic €64.99
Hybrid search

The Best of Keywords and AI in One Search

Hybrid search runs keyword matching and semantic search in parallel, then merges the results. Keywords handle exact matches, model numbers, and SKUs. Semantic embeddings handle meaning — connecting “running gear” to “athletic shoes” even when the words do not overlap. You control the balance with a single slider.

OpenAI embeddings — choose from text-embedding-3-small, text-embedding-3-large, or ada-002
Configurable semantic weight — slide from 0.0 (pure keyword) to 1.0 (pure semantic), default 0.5
Client-side or server-side — generate embeddings in PHP or let Meilisearch handle them automatically
Best of both worlds — exact matches still rank high, but meaning-based results fill the gaps
Best for: Stores where customers describe products in their own words rather than searching by product name or SKU.
Hybrid Search
Keyword Semantic
Semantic Weight: 0.5
Keyword results
Running Shoe Pro X — exact name match
Marathon Racer 3000 — contains “running”
Semantic results
Athletic Trail Trainer — meaning match
Sprint Performance Sneaker — intent match
Hybrid results (combined)
1. Running Shoe Pro X
2. Athletic Trail Trainer
3. Marathon Racer 3000
4. Sprint Performance Sneaker
Query expansion

One Search Term, Ten Times the Reach

AI automatically expands search terms with synonyms and related phrases. When a customer searches “sneakers,” the module also searches for “trainers,” “running shoes,” and “athletic shoes.” No manual synonym lists. No maintenance. The AI handles vocabulary gaps so your catalog speaks every customer’s language.

Automatic synonyms — “sneakers” also searches “trainers,” “running shoes,” “athletic shoes”
Regional vocabulary — bridges “couch” to “sofa,” “trainers” to “sneakers,” “mobile” to “cell phone”
Zero maintenance — AI generates expansions on the fly, no manual synonym lists needed
Catalog-aware — expansions are grounded in your actual product data, not generic dictionaries
Query Expansion
Customer searches:
sneakers
AI expands →
Module searches for:
sneakers trainers running shoes athletic shoes sport shoes
32 products found (vs. 8 with keyword only)
Natural language

Customers Ask Questions. Your Search Answers Them.

Customers type the way they talk: “what’s good for running?” or “something to keep drinks cold at the beach.” Traditional search chokes on these queries. Ultrafast Search converts natural language questions into precise keyword searches — automatically. No training needed, no rules to configure.

Questions become searches — “what’s good for running?” finds running shoes
Intent recognition — understands the purpose behind the words, not just the words themselves
Descriptive queries — “something warm for winter hiking” finds insulated hiking boots
Conversational search — customers search like they speak, and your store understands
Natural Language Search
What’s a good gift for someone who likes cooking?
AI understanding intent...
cooking gifts, kitchen accessories, cookware sets
Professional Chef Knife Set €89.99
Italian Ceramic Cookbook Stand €34.50
Bamboo Cutting Board Collection €42.00
Smart fallback

Zero Results? Not Anymore.

When keyword search returns nothing, most stores show an empty page. The customer leaves. With Ultrafast Search, AI fallback activates automatically — re-interpreting the query, broadening terms, and surfacing relevant products. The “no results” dead end is replaced with a second chance at the sale.

Automatic fallback — triggers instantly when keyword search returns zero results
AI re-interpretation — broadens or rephrases the query to find related products
Never a dead end — customers always see relevant suggestions instead of an empty page
Reduces bounce rate — keeps shoppers engaged when their exact terms do not match
Smart Fallback
Without AI fallback
ergonomic desk chair for back pain
0 results found
AI fallback activated
With smart fallback
ergonomic desk chair for back pain
AI-assisted results
ErgoMax Lumbar Support Chair €349.00
PosturePro Office Chair €289.99
BackCare Mesh Desk Chair €199.00
Ergonomic Seat Cushion €49.99
Embedding models

Three OpenAI Models. You Choose.

Different stores need different levels of precision. Choose the embedding model that fits your catalog size and budget. Then configure exactly which product fields generate embeddings — product name, description, category, manufacturer, reference, or any combination. You control what the AI learns about your products.

text-embedding-3-small — fast and cost-effective, ideal for most stores
text-embedding-3-large — highest accuracy for large or complex catalogs
ada-002 — proven legacy model, stable and reliable
Per-field control — embed name, description, category, manufacturer, reference, or any combination
AI Embedding Settings
Embedding Model
text-embedding-3-small
text-embedding-3-small (recommended)
text-embedding-3-large (highest accuracy)
ada-002 (legacy, stable)
Embedding Mode
Client-side (PHP)
Fields to Embed
Product name
Description
Short description
Category name
Manufacturer name
Reference
Semantic Weight: 0.5
0.0 1.0
Cost control

AI Search You Can Actually Afford

AI sounds expensive. It does not have to be. Ultrafast Search includes three built-in cost controls: an AI response cache that eliminates repeat API calls, batch embedding generation that processes 50–100 products per single API call, and token monitoring that tracks every cent in USD. Your OpenAI API key is the only external cost — and the module minimizes it aggressively.

AI response cache — repeat queries served from cache, zero API calls needed
Batch embeddings — 50–100 products per API call, 10–50x faster than one-by-one generation
Token monitoring — track usage and cost in USD per day, week, or month
Cache hit/miss tracking — see exactly how much the cache saves you
All AI features are included in your subscription. No per-query AI surcharges, ever. See pricing.
AI Cost Dashboard
Cache Hit Rate 87%
Tokens Used Today 12,450
Est. Daily Cost $0.03
Batch Speed 50 products/call
87 of 100 queries served from cache
How it works

From Question to Product in Five Steps

The entire AI search pipeline is automatic. No machine learning expertise needed — just enable AI in your admin panel and your search gets smarter instantly.

1

Customer types

A natural language query, question, or keyword in the search bar

2

AI expands the query

Synonyms added, questions converted to keywords, terms broadened

3

Hybrid search runs

Keyword matching + semantic embedding search execute in parallel

4

Results ranked

Combined scoring from keyword relevance and semantic similarity

5

Results displayed

Products appear in your chosen display mode

5 display modes →

Experience AI Search for Your Store

Try AI-powered semantic search, query expansion, and natural language understanding free for 14 days. No credit card required.

Start Free Trial → Explore All Features →