Search API: Similar Products
Product Search
Search API: Similar Products
Similar Products API endpoint that returns visually and semantically related items to a given product using image and text embeddings from your catalog.
POST
Search API: Similar Products
How results are served
Similar products results are precomputed and cached for fast response times. The system runs the full ranking pipeline — combining vector similarity with behavioral signals — and stores the top results for each product. When you call this endpoint, results are served from the precomputed cache. If no precomputed data exists for a product (for example, a newly added product), the system falls back to real-time vector search automatically. Precomputed results are refreshed daily and incrementally when a product’s embeddings change. All request parameters (filters, facets, pagination, and sort orders) work the same regardless of whether results are served from cache or computed in real time.Authorization
Token-based authentication header in the form of
<YOUR_LAYERS_TOKEN>.Headers
Path parameters
The ID of the product for which you want to find similar products.
Body
Product attributes to include in the response. By default, all attributes are included. Available attributes include:
id, title, handle, body_html, vendor, product_type, tags, images, available, created_at, updated_at, published_at, price_range, options, original_options, metafields, named_tags, calculated, category, featured_media, is_gift_card, has_variants_that_require_components, combined_listing_parent_product_id, combined_listing_role, first_or_matched_variant, and variants. See the Product Schema for detailed descriptions, including nested field selection for slicing arrays and filtering collections (for example, images[:2].src, variants[sku=ABC].price, metafields[namespace=custom]).Refer to our dedicated Filter Expressions guide to learn more about filter expressions.
Facets to be included. Accepts both exact facet codes (e.g.,
"vendor", "options.Size") and wildcard patterns (e.g., "options.*", "metafields.product.*").Wildcard patterns expand to all matching attribute codes. For example, "options.*" expands to all option facets like "options.Size" and "options.Color". Wildcards must match at least one attribute code to be valid.Examples:If the count of each facet value should be calculated
If you want a min/max range for numeric facets such as price.
Code of the sort order to apply. Use
"relevance" for the default relevance-based ranking, or any custom sort order code configured in your store. Retrieve available sort order codes from the Sort Orders API.Specify which product variants should be selected by default in the
first_or_matched_variant field.Behavior:- Uses OR logic - selects the first variant matching ANY of the specified option values
- Falls back to the first variant by position if no match is found
- Automatically skipped when any option or variant filters are applied (filters take precedence)
- Only affects the
first_or_matched_variantfield in the response
Apply discounts to products, variants, or collections. When provided, price-based sorting and filtering will use discounted prices. See the Discount Entitlements guide for detailed usage.
User identity information for tracking and personalization. Automatically managed by the Storefront Pixel; required for headless integrations.
Contextual information about the customer’s session, behavior, and environment. Automatically collected by the Storefront Pixel; must be manually provided for headless integrations. See Contextual Information for implementation guidance.
Per-request overrides for response behavior. Currently controls image preload hints emitted in the
Link response header. You can disable preloads or narrow the srcset to the breakpoints your storefront uses. You cannot enable preloads when the store has them disabled.Example:Response
The total number of results.
The current page number.
The total number of pages.
If
includeFacetRanges is true then an object with keys of the facet attribute code and value is an object with min/max.If
retrieveFacetCount is true then an object with keys and values of the specified attributes. Empty values and literal "null" strings are automatically excluded from facet results.A unique request identifier (ULID) for this API call. Use this token to correlate requests with analytics events via the Beacon API.
An ordered trace of processing steps the engine performed. See the Text Search documentation for the full list of possible step types.
Response headers
| Header | Description |
|---|---|
x-request-id | A unique identifier (ULID) for the request. This matches the attributionToken in the response body. |
x-layers-build | The Layers build version that served the request. |
X-Layers-Results-Cache | Indicates whether the response was served from the results cache. Returns HIT or MISS. |
Link | Contains rel="preload" directives for product images (up to 6 items). See Early hints for details. Only included when early hints are enabled (enabled by default). |