eCommerce

AI product search on Shopify, Magento and WooCommerce: options compared

13 min readBy Jaskirat Singh Talwar

As of October 2026, every major store platform offers a form of AI-assisted product search, and the real differences are in what your licence already includes, what you can tune, and whether the option exists for your platform at all. Shopify switches on semantic understanding automatically for eligible stores; Adobe Commerce includes Live Search in the licence but Magento Open Source is not on its supported list; WooCommerce has no native equivalent and relies on extensions or third-party engines. Third-party engines work across platforms on request-based or quoted pricing, and a build-your-own path (embeddings plus pgvector or OpenSearch) makes sense for stores with developers and an unusual catalogue. This is a sourced comparison from vendor documentation, not a hands-on test; what you measure on your own search logs decides it.

Key takeaways

  • "AI search" covers four different things: typo and synonym handling, semantic (meaning-based) matching, AI ranking and personalisation, and merchandising rules. Ask a vendor which of the four they mean.
  • Shopify's semantic search applies automatically to online store search for stores with fewer than 200,000 products on the Grow, Advanced or Plus plans; it does not apply to predictive search or the Japanese locale.
  • Adobe Commerce Live Search is a SaaS service "included in your license" for Adobe Commerce 2.4.4 and later; it cannot run alongside OpenSearch on the same instance, and Adobe's requirements page lists only Adobe Commerce on Cloud and on-premises.
  • Third-party engines price by search requests and records (Algolia, Doofinder, Jetpack Search) or by quote (Athos Commerce, the January 2025 combination of Klevu and Searchspring). Link to the pricing pages; the numbers change.
  • The decision metrics are the same whatever you choose: zero-result rate, search click rate, search-to-purchase rate and revenue per search, measured before and after.

What "AI search" means in 2026

Vendors use one label for several capabilities. Separate them before comparing anything:

  1. Lexical search with tolerance. Keyword matching plus typo correction, stemming and synonyms. Every option here does this; the quality differences are in the synonym tooling.
  2. Semantic search. The query and the products are converted into embeddings (numeric representations of meaning) so "something warm for a winter run" can match a thermal running jacket that never uses those words. Shopify's help documentation describes its version as using "product attributes such as descriptions and image data, including text within images and colors".
  3. AI ranking and personalisation. Re-ordering results from behaviour: Adobe describes Live Search as "AI-powered dynamic faceting and re-ranking of search results in response to in-session shopper behaviors"; Algolia sells AI Ranking and Advanced Personalization on its Grow Plus plan and NeuralSearch only on its Elevate plan.
  4. Merchandising rules. If-then boosting and burying, pinned products, campaign rules. Not AI, but it is what merchandisers actually use day to day.

Most stores get the biggest lift from 1 and 4 done properly; 2 and 3 matter more as the catalogue grows and as shoppers type in natural language, which the shift to AI assistants is training them to do (see our guide to agentic commerce readiness).

Native options, platform by platform

Shopify

Shopify's own Search & Discovery app is free, built by Shopify, and provides filters, synonym groups, product boosts, related and complementary product recommendations, and search analytics. Its App Store rating was 2.6 out of 5 from 508 ratings in October 2026, with a split between one-star and five-star reviews, so read the reviews for your use case before relying on it.

Semantic understanding sits underneath: Shopify's help documentation says semantic search is available for online store search when "your store has fewer than 200,000 products" and is on a Grow, Advanced or Plus plan, that "you don't need to turn it on in the Search and Discovery app", that it "doesn't apply to predictive search" and "isn't supported for the Japanese locale". If your theme or an app generates its own results, you are not using it.

Shopify's search performance reports (last 30 days by default in the app, longer ranges under Analytics) cover click rate, purchase rate, searches by query, searches with no results and searches with no clicks. That is enough to run the evaluation procedure below without buying anything.

Adobe Commerce (Magento)

Live Search replaces the default search and adds a product listing widget with facets. Adobe's documentation describes it as a SaaS service "included in your license" with "AI-powered dynamic faceting and re-ranking", a synonym dictionary, and merchandising rules. The requirements page lists "Adobe Commerce 2.4.4+" on Cloud and on-premises; Magento Open Source does not appear. Two implementation facts matter: "OpenSearch and Live Search cannot both be enabled on the same Commerce instance", and the catalogue is synced to Adobe's SaaS infrastructure (the instance needs access to commerce.adobe.io, running cron and "Update by Schedule" indexers).

Published limits, as of the current Boundaries and Limits page: 450 indexed product attributes per store view (50 sortable, 200 filterable, 200 searchable), up to 100 facets, 200 synonyms per store view, 50 search merchandising rules, and pagination to 10,000 results at 100 per page. Performance metrics are kept for one year.

A Magento Open Source store therefore has two routes: OpenSearch with extensions for synonyms and merchandising, or a third-party engine. Our refreshed Magento and Adobe Commerce guide covers the upgrade path.

WooCommerce

WooCommerce core search is keyword matching over the WordPress database, with no semantic layer, so the options are extensions:

  • AI Semantic Search (sold on woocommerce.com) connects the store to OpenAI with your own API key "required for generating product vectors", offers Semantic, Keyword and Hybrid modes, and lets you choose which fields are indexed (title, SKU, descriptions, categories, tags, price text, attributes). You pay the extension subscription plus OpenAI usage.
  • Jetpack Search (Automattic) prices by records and monthly requests; the free tier covers 5,000 records and 500 requests a month, with instant search, filtering and spelling correction, and an AI Answers feature on paid plans.
  • Algolia for WordPress and WooCommerce runs through the "WP Search with Algolia" plugins by WebDevStudios, which Algolia's documentation points to; the Pro version indexes product data including SKU, pricing and variations.
  • Doofinder and other hosted engines list WooCommerce among their integrations.

See our WooCommerce guide for the surrounding performance and checkout decisions.

Third-party engines across platforms

Option Platforms (per vendor docs, Oct 2026) How the "AI" works Pricing model Tuning controls Best fit
Shopify Search & Discovery + semantic search Shopify only Semantic matching on eligible plans; boosts and synonyms Included; plan-gated (Grow/Advanced/Plus, under 200k products) Synonyms, boosts, filters Shopify stores that have not yet measured search
Adobe Commerce Live Search Adobe Commerce 2.4.4+ (Cloud, on-prem) Dynamic facets, behaviour-based re-ranking, synonyms, rules Included in licence 200 synonyms, 50 search rules, facets Adobe Commerce stores; not Open Source
WooCommerce AI Semantic Search extension WooCommerce OpenAI embeddings, hybrid mode Subscription + your OpenAI usage Field selection, mode Small to mid WooCommerce catalogues
Algolia Shopify, Adobe Commerce/Magento 2, BigCommerce; WooCommerce via WebDevStudios plugin Keyword with AI Ranking/Synonyms (Grow Plus); NeuralSearch (Elevate) Per search request and per record; free tier 10,000 requests and 50,000 records; Elevate on annual contract Rules, ranking, merchandising, personalisation Multi-platform or headless builds with developers
Athos Commerce (Klevu + Searchspring) BigCommerce, Miva, Salesforce, Shopify, Shopify Plus, Adobe Commerce, Shopware, NetSuite (WooCommerce not listed) AI search, merchandising, recommendations, feed management Quote ("Request Pricing") Merchandising suite Mid-market stores wanting search plus merchandising in one vendor
Doofinder Shopify, WooCommerce, Adobe, PrestaShop, BigCommerce AI search, visual search, personalisation Tiers by monthly search requests (10k, 150k, 400k, custom) Merchandising, suggestions Stores on several platforms wanting low-setup hosted search
Build your own Any Embeddings + vector/hybrid index you control Infrastructure and engineering time Everything, and you maintain it Unusual catalogues, strict data control, existing dev team

Two notes on the vendors. Klevu and Searchspring combined in January 2025 under the Athos Commerce brand, so older comparisons that list them separately are out of date. Algolia's NeuralSearch (its semantic product) is only on the Elevate plan with an annual contract; its Free and Grow plans are keyword search with AI features added at Grow Plus, which is easy to miss on a feature grid.

Building your own

A store with developers and an unusual catalogue (industrial parts, compatibility-driven products, multilingual ranges) can build search from open components:

  • Embeddings. A model such as Amazon Titan Text Embeddings V2 takes up to 8,192 tokens and returns a 1,024-dimension vector (512 and 256 also available); it is "optimized for English, with multilingual support" for 100-plus languages in preview. Product titles, attributes and descriptions are embedded once and re-embedded on change.
  • Index. pgvector adds "open-source vector similarity search for Postgres" with HNSW and IVFFlat indexes (vectors up to 16,000 dimensions; indexed vector columns up to 2,000). OpenSearch's hybrid query lets you "combine relevance scores from multiple queries into one score for a given document", so keyword and vector scores are merged through a search pipeline.
  • Hybrid by default. Pure vector search is bad at exact SKUs and model numbers; pure keyword search is bad at intent. Combine them from day one, as the vendors above do internally.

Cost is mostly engineering and ongoing ownership; the embedding and hosting bills are small for a catalogue of tens of thousands of products, but check the provider's current pricing page rather than any article. This path is only worth it when the hosted options genuinely cannot express your catalogue's logic. We have built embeddings-based search of this kind; in our experience the index is the quick part, and keeping the product data clean and the evaluation queries current is the ongoing work.

What to measure

Define four numbers before touching anything, from your existing reports:

  1. Zero-result rate: searches that return nothing, divided by all searches. Shopify's "Searches with no results" report gives it directly.
  2. Search click rate: searches where the shopper clicks a result. Shopify reports it; Live Search tracks performance metrics for the last year.
  3. Search-to-purchase rate: Shopify's "purchase rate" report; in other stacks, sessions with a search that end in an order, divided by sessions with a search.
  4. Revenue per search: revenue from searching sessions divided by searches. This is the number that justifies a subscription.

Track them weekly on a shared sheet; a search change that does not move at least one of them within four weeks has not worked.

Implementation gotchas

  • Catalogue quality beats engine choice. Semantic search embeds what you wrote. Products with one-line titles and empty attributes retrieve badly in every engine.
  • Synonyms are still needed. Brand nicknames, regional words and internal codes do not emerge from embeddings; every vendor here still ships a synonym tool, and Live Search caps it at 200 per store view.
  • Language. Shopify excludes the Japanese locale from semantic search; Titan Text Embeddings V2 is English-optimised with other languages in preview. Test non-English queries explicitly.
  • Predictive search versus the results page. On Shopify, semantic understanding applies to the results page, not to the predictive dropdown; report on both separately.
  • Indexing lag. Live Search depends on cron and scheduled indexers; a price change that reaches search an hour late is a support ticket.
  • Theme and app conflicts. A theme or app that renders its own results silently bypasses the platform's semantic layer.
  • Measurement first. Without the four numbers above from before the change, you will not know whether the subscription paid for itself.

A six-step evaluation on your own store

  1. Export the last 90 days of search terms with counts, zero-result flags and clicks (Shopify's reports; Live Search performance; server logs or a search plugin for WooCommerce).
  2. Compute the four metrics and list the top 200 queries by volume plus the top 50 zero-result queries.
  3. Shortlist two options from the table that exist for your platform and plan.
  4. Run the 250 queries through each option on a staging store and score the top five results as relevant, partly relevant or wrong, with one merchandiser and one developer scoring independently.
  5. Put the winner live for four weeks and compare the four metrics with the prior four weeks, same promotions where possible.
  6. Decide on the numbers, and record why, so the next review in twelve months starts from evidence.

FAQ

Does Shopify have AI product search built in?

Yes, for eligible stores. Shopify's documentation says semantic search applies to online store search when the store has fewer than 200,000 products and is on the Grow, Advanced or Plus plan; it is on automatically, does not apply to predictive search and is not supported for the Japanese locale. The free Search & Discovery app adds synonyms, boosts, filters and reports.

Is Adobe Commerce Live Search available for Magento Open Source?

Adobe's requirements page lists Adobe Commerce 2.4.4 and later on Cloud and on-premises only, and describes Live Search as included in the Adobe Commerce licence. Magento Open Source stores use OpenSearch with extensions, or a third-party engine such as Algolia, Athos Commerce or Doofinder.

What is the best AI search plugin for WooCommerce?

There is no single answer; it depends on catalogue size and budget. The AI Semantic Search extension on woocommerce.com uses your own OpenAI key with a hybrid mode; Jetpack Search prices by records and requests with a free tier; Algolia works through the WebDevStudios plugins; Doofinder lists WooCommerce. Measure zero-result rate and purchase rate before and after any of them.

How much does AI site search cost?

It depends on the pricing model: included in the platform (Shopify semantic search, Live Search), per request and record (Algolia, Doofinder, Jetpack Search), by quote (Athos Commerce), or engineering time plus small infrastructure costs if you build it. Prices change often; use the vendors' current pricing pages rather than figures from articles.

Should we build our own semantic search?

Only if a hosted option cannot express your catalogue's logic and you have developers to maintain it. The components are mature (embedding models, pgvector, OpenSearch hybrid queries), but ownership of relevance tuning, synonyms and monitoring stays with you for the life of the store.

Impinge Solutions builds and maintains Shopify, Adobe Commerce and WooCommerce stores and offers a store AI readiness check that starts with your search reports. Ask for the search evaluation on your store.

Sources

Are you ready to partner with us?

Our good-natured team is waiting to hear your business requirements.

Book a free project call

Stay informed with our newsletter