How to fix product schema for AI search.

How people do this without eCommerce Insights

The manual approach: you open one PDP, view source, copy the JSON-LD block into a validator, look at what's missing, and write a one-off fix. Then you do it again for another PDP. By SKU forty you realize this will never finish, so you hand it to a developer. The developer starts looking at theme liquid, discovers the JSON-LD is partially built from metafields and partially hard-coded, and estimates three weeks.

Alternative: you install a JSON-LD Shopify app. It adds a more elaborate schema block. You check three products and they look better. You ship. Two weeks later the Shopify SKU Visibility Grader shows your citation rate did not move, because the app's schema is technically valid but structurally shallow — it fills the easy fields and ignores the metafield-backed ones that matter. You are back where you started.

The gap is that schema is not just a technical compliance problem. It is a content problem. The material, care, origin, and fabric fields come from the brand's knowledge of its own catalog; no generic theme or app can fill them. You need a tool that knows your Shopify metafield schema, validates per field, and shows the content owner what is missing at the catalog level.

How to do this in eCommerce Insights

1. Audit current schema across the catalog. Run the eCommerce Insights schema audit (one click once your Shopify store is connected). eCommerce Insights fetches every PDP, parses its live Product JSON-LD, validates against schema.org and the AI-engine field expectations derived from Q1 2026 citation data, and scores each SKU.
2. Review the field-level gap report. The gap report shows which fields are missing per SKU, sorted by fix impact. It groups gaps by type (aggregateRating missing on X SKUs, material missing on Y SKUs) so you can batch-fix rather than working one product at a time.
3. Generate complete JSON-LD per SKU. eCommerce Insights drafts the replacement JSON-LD using metafield data, product JSON, and review app data. Every draft goes through field-level validation. Fields eCommerce Insights cannot fill automatically (brand-specific values like origin or fiber blend) are flagged for you to fill via the Shopify metafield editor.
4. Approve diffs. Each SKU's proposed schema arrives as a diff against the live version. Approve individually for hero SKUs or in batch for long-tail. Reject any draft that needs manual editing. The approval queue is the whole merchandising team's UI — a PM can review the diffs without reading JSON.
5. Push to Shopify. Approved schema is written to Shopify metafields in a eCommerce Insights namespace. Your theme references those metafields in the PDP JSON-LD block. No theme-file edits after initial setup. This is early access; the manual alternative is to export the schema as copy-paste snippets for theme-file installation.

For the underlying schema mechanics, see the Schema for AI search guide.

What "good" looks like

• Eleven core fields present on 95%+ of PDPs: name, description, sku, gtin, brand, offers (price, availability, url), image, material, color, aggregateRating, review.
• No PDPs scoring below 60 on the structured-data factor. A score below 60 usually means two or more core fields are missing.
• aggregateRating populated wherever review data exists. The single most common miss — review data exists in the store's review app but never reaches the JSON-LD block.
• Zero schema validation errors in Google's Rich Results Test for hero SKUs. Warnings are acceptable; errors mean fields like price or availability are malformed.

Related jobs

Related tools

• Product schema generator — free per-SKU schema generator with field-level validation.
• AEO Grader — store-wide readiness score with structured-data breakdown.