Predictive search on Shopify: what it returns, how it matches, and how to make your catalog match it

Predictive search on Shopify is the dropdown that suggests results while a shopper is still typing in the search box. According to Shopify's developer documentation, checked on 12 October 2026, it can suggest products, collections, queries, pages and articles, and it matches search terms exactly or with typo tolerance against the searchable properties of your shop. So the way to improve what it shows is to put the words shoppers type into the product fields it reads: title, product type, tags, variant titles and description.

What the dropdown contains

The Shopify guide to adding predictive search to a theme describes a dropdown with five kinds of suggestion: queries, collections, products, pages and articles. Query suggestions are generated by extracting words and phrases from your product catalog and from customer searches, using natural language processing. Product suggestions are the products or variants that match the typed terms.

Shopify's own example is a shopper who types "very fast snowbo". Product suggestions appear for items that contain the words "very" and "fast" and a word that begins with "snowbo". Each typed word narrows the list, which means a product that is missing one of those words in its fields will not appear.

The rules that decide a match

Rule	What the documentation says	What to do about it
Typo tolerance	Results can differ from the search term by one letter, or have two letters in a different order. The first four letters of the term need to be correct.	Spell product names the way shoppers spell them, and do not rely on typo tolerance for the first letters
Word breaks	A word separated by a hyphen or a space counts as two terms. The example given: T-shirt and t shirt return the same results, but tshirt does not.	Use the spelling shoppers type, and add the variant spelling to a tag or the description
Product fields	Product results search title, body, product type, tag, vendor, variant title, variant SKU, variant barcode and product category	Fill these fields with the real words, not internal codes only
Variants	A variant is returned only when the query matches terms specific to the variant title, and only the variants with the most matching terms are returned	Give variants meaningful titles, such as colour and size words
Language	Predictive search is supported only when the buyer locale is one of the supported languages	Check the language list in the guide before you promise it to a market

The documentation also notes that predictive search uses a different search engine from the main storefront search, so it does not handle partial word matches in the same way. If your shoppers use both, test both.

The settings you control

The API reference lists default behaviour you can change through Search Settings in the Search and Discovery app. Those documented defaults, as at the check date, are:

• Result types returned: query, product, collection and page.
• Maximum results: a value from 1 to 10, with 10 as the default.
• Unavailable products: shown below the other matching results by default, and can be shown or hidden.

Decide on the unavailable setting deliberately. A sold out item at the top of the dropdown loses the shopper's attention, and hiding it can lose the shopper who would have joined a waiting list. Showing it below the available matches, the documented default, is a sound middle position.

How the API works if you build your own

A developer can call the Predictive Search API with a request in the form GET /{locale}/search/suggest.json?q={query}. The query parameter q is required. Optional parameters set the type of resource (resources[type]), the number of results (resources[limit]) and how that limit is shared across types (resources[limit_scope], with all or each). Shopify says requests should use locale aware URLs so visitors get a consistent experience. The reference page is the Predictive Search API reference, and it includes a short example that reads the first product suggestion from the JSON response.

The theme guide adds a note on accessibility: its examples follow the listbox component pattern from the W3C ARIA authoring practices. Keep that pattern if you change the markup, so keyboard and screen reader users can move through suggestions.

Eight catalog edits that improve what appears

1. Write titles with the product's plain name first. "Leather weekender bag, tan" matches more typed searches than a brand coined name alone.
2. Fill in product type. It is a searched field, and it is the word a shopper uses for a category.
3. Use tags for the other words people use. Add "holdall" or "duffel" as a tag if shoppers call a weekender bag by those names.
4. Add spelling variants. Given the hyphen and space rule, put both "t shirt" and "tee" somewhere in the fields.
5. Name variants in words. A variant titled "Light blue / Large" can be found by a search for "light blue", as in Shopify's own snowboard example.
6. Keep SKUs and barcodes in the variant fields. Trade buyers search by SKU, and the field is searched.
7. Use collection titles that match category searches. Collections are suggested too.
8. Publish pages and articles for the questions that are not product names, such as sizing, delivery and returns. Pages and articles appear in the dropdown when they match.

How to test it

Open your storefront and type twenty searches you would expect real shoppers to use. Include a misspelling, a two word phrase with and without a space, a SKU, a colour and size combination, and a question such as "returns". Note what comes first for each and what is missing. Fix the field that explains each gap, and repeat the same twenty searches after the edit. Keep the list so you can run it again after every catalog import.

Site search is one part of store discovery. The wider checklist, including collection pages and variant URLs, is in Shopify SEO, and the full set of ecommerce mechanics is in ecommerce SEO.

Questions about Shopify predictive search

Is predictive search the same as the search results page?

No. The dropdown appears while typing and is served by the Predictive Search API. The results page is the storefront search. Shopify documents that they use different engines and handle partial words differently.

Does predictive search work in every language?

It works when the buyer's session is in one of the languages Shopify lists as supported. Read the Requirements and limitations section of the API reference for the current list.

Can I show only products in the dropdown?

Yes. The resources[type] parameter accepts a comma separated list of product, page, article, collection and query, so you can request only the type you want.

Why does a product with a typo in the search still appear?

Typo tolerance is set to one letter, so a term that differs by one letter, or has two letters swapped, can still match, provided the first four letters are correct.

Where do I change the default result types?

Shopify's reference says the defaults can be changed using Search Settings in the Search and Discovery app.

The shopper who skips the search box and asks an assistant

Predictive search serves the shopper who is already in your store. A growing group of shoppers asks an assistant first, and the assistant answers from product data and pages around the web, not from your dropdown. The same edits help in both places, because a clear title, product type and plain description are what the dropdown matches and what an assistant reads. To see how your own products are described outside the store, ask ChatGPT, Gemini, Perplexity and Claude for the best option in your category and read who is named. The page on Shopify and ChatGPT explains the routes in, and AI Knows Us (aiknowsus.com) can run the question set for a store and show the sources behind each answer.