Shopify Product Recommendations Not Working? Diagnose Empty, Irrelevant, or Missing Recommendations

The direct answer

Shopify recommendations are usually broken by a path, not a single setting

A recommendation has to travel from a product context to a candidate source, through eligibility rules, into a theme or API response, through a product card, and finally into an analytics event. If any link fails, the shopper sees an empty row, the wrong products, no products, or a product that cannot be purchased.

That is why changing the heading or selecting more products rarely fixes the real problem. First ask whether the module is present. Then ask whether Shopify is allowed to return the candidate. Then ask whether the candidate answers the page’s question. Only after that should you tune ordering or presentation.

Layer 1

Check theme support and placement first

Shopify product recommendations do not appear on every product template automatically. Shopify documents that your theme needs the relevant product recommendation or complementary product section, and that the section needs to be added to the product page. If the section is absent, candidate quality is irrelevant because there is no visible surface.

Open the theme editor and inspect the exact product template used by the anchor. Stores often have more than one template. A recommendation section may exist on the default template but not on a custom template for bundles, wholesale products, landing pages, or seasonal products.

If the module appears on one product and not another, compare templates before comparing products. If it appears nowhere, inspect theme support and section configuration before touching Search & Discovery.

For a custom storefront, confirm that the recommendation query runs after the anchor product is known, that the request uses a published product identifier, and that the response is inserted into the page.

Layer 2

Check product eligibility before judging relevance

Shopify applies product eligibility rules before a recommendation reaches the shopper. The candidate must be active, published to the Online Store channel, and priced above zero. It cannot be a gift card or an unlisted product. The inventory rule differs by intent: related products can have a documented exception when continue selling is active, while complementary products require inventory above zero.

The candidate also cannot already be in the visitor’s cart. This is easy to mistake for an algorithm problem. A shopper adds an accessory, revisits the product page, and the accessory disappears from the complementary row. That may be the correct behavior, not a missing recommendation.

Check the candidate in the admin and on the storefront. A product can be active in the admin and still be unpublished to the Online Store channel. A product can be published and still fail the card because its chosen variant is unavailable. Record the exact state you see.

Check the anchor too. The recommendation request can fail because the anchor product is not found, is not published, or is identified incorrectly in a custom integration. A missing anchor is not the same as an empty candidate set.

Layer 3

Identify whether the source is manual, automatic, or custom

A Shopify store can show recommendations from manual selections, automatic related-product logic, a theme or API implementation, or another app. Those sources can look identical to a shopper but behave differently when data changes.

Open Search & Discovery and inspect the anchor product. If you selected manual recommendations, verify that the list is still intentional. If you enabled automatic recommendations, understand that the output can change as product data, orders, and customer activity change. Shopify documents purchase history, product descriptions, and related collections as automatic recommendation inputs, with different availability depending on the store.

The right source can vary by category. Use automatic relationships for a broad catalog with enough activity. Use manual relationships for parts, compatibility, bundles, and strategic launches. Use a hybrid approach when the store needs both coverage and exceptions.

Layer 4

Audit catalog data and cold-start conditions

Automatic recommendations need evidence. New products, low-volume products, imported orders, and sparse descriptions can produce thin or surprising relationships. This does not necessarily mean the system is broken. It means the store needs an intentional fallback while the product earns enough context.

Inspect the anchor and candidate titles, descriptions, product types, collections, tags, attributes, and compatibility fields. Ask whether a human can explain the relationship from the data. If the answer is no, automatic logic has less to work with and manual selection will be harder to govern too.

Be careful with imported history. If order history came from another store or platform, it may not support the same recommendation behavior as new orders in the current store. For a new product, use the fields and relationships you know rather than assuming the system can infer a complete network from day one.

Then check the candidate card. A recommendation can be logically correct but look irrelevant because the title is generic, the image is misleading, or the variant detail is hidden. Data quality and presentation quality reinforce each other.

Layer 5

Trace the API response and card rendering separately

If the candidate set is correct but the storefront is wrong, inspect the boundary between data and presentation. Shopify’s Ajax Product Recommendations API can return product data or a rendered section. A custom theme or headless implementation can then alter, filter, cache, or replace that response before the shopper sees it.

Use the browser’s network panel or storefront logs. Confirm the request contains the correct product ID, limit, intent, locale, and section ID where relevant. Confirm the response contains the expected product. Then inspect the DOM and verify that the card reads the same product, price, availability, URL, and image.

recommendations/products.json
 product_id = anchor product
 intent = related or complementary
 limit = the number your card can display
 locale = the shopper’s storefront locale
 compare response → rendered product → clicked URL → analytics event

A 404 can mean the product does not exist or is not published to the Online Store channel. A 422 can mean a required parameter is missing or the intent is invalid. Shopify documents both response patterns in the Ajax API reference. Use the response code to choose the next owner instead of treating every empty module as a ranking issue.

Layer 6

Verify measurement before you call the fix a success

A recommendation can look fixed in the browser and still be invisible in analytics. Shopify documents recommendation click, add-to-cart, and purchase reporting, but the implementation needs to preserve the tracking path. A custom card that strips recommendation parameters can make the dashboard undercount or misattribute activity.

Test one anchor and one candidate. Open the module, click the candidate, add it to the cart, and complete a test purchase only if your environment allows it. At minimum, inspect the URL, event payload, product ID, recommendation intent, placement, and session ID. Record what you can prove.

If the module is visible but the report is empty, fix measurement before changing candidates. If the report has clicks but the storefront test cannot reproduce them, check event scope, consent behavior, URL parameters, and whether another recommendation surface is mixed into the same report.

Layer 7

Run the 30-minute recovery audit

Use this sequence when a merchant report says “recommendations are not working.” It gives you evidence without requiring a complete implementation review.

Frequently asked questions