The structured data your AI competitors already have

The Gap Between Cited and Invisible

When an assistant names a product, compare the answer and any cited sources with the product record. Check the name, price, availability and other stated facts. Do not assume every detail came from JSON-LD or that missing markup explains an omitted product.

Structured data describes product facts for systems that consume it. For Google AI Overviews and AI Mode, Google’s guidance requires no special AI schema. Supported markup can establish eligibility for particular Google rich results; it does not guarantee a citation, ranking or recommendation.

Most commerce sites have some structured data. Very few have structured data that’s detailed enough to compete for AI citations.

Source: Google Search: AI features and your website

Choose Markup for the Destination

For Google Search, distinguish product snippets from merchant listings. A page where customers can buy a product should follow the merchant-listing requirements; other product pages may use product-snippet requirements. Supply the required properties for that feature and applicable supported details from the actual product record. Do not treat every desirable product attribute as a required or recognized Product property.

Keep useful product questions and verified answers visible on the page. Google stopped showing FAQ rich results on May 7, 2026. This is a Google feature change, not the disappearance of the FAQPage vocabulary or a rule for every other consumer. Do not add FAQPage markup expecting a Google FAQ result or guaranteed AI citation.

Use Review or AggregateRating markup only for genuine review content that meets the selected feature’s rules and matches what readers can see. Check the rating value, review count and source records. Do not invent ratings to fill a field or assume that a higher review count guarantees a recommendation.

Publish clear, supported use instructions where they help the shopper. Google deprecated HowTo rich results in September 2023. A readable sequence can still be useful content, but HowTo markup is not a current Google rich-result requirement or a guarantee that an assistant will surface the instructions.

Source: Google Search: Product structured data

Source: Google Search: FAQ rich-result deprecation

Source: Google Search: HowTo rich-result deprecation

Source: Google Search: Structured data guidelines

How to Audit What You Have

Run representative product URLs through Google’s Rich Results Test. It reports supported Google rich-result types and their errors or warnings; it is not an inventory of every schema.org type or an AI-readiness certification. Inspect the rendered page and its markup as well.

Inspect what the theme and installed apps actually emit. Check the product identity, offer and variant values against the visible page and source records. For any review markup, verify that the rating and count match real reviews. Review useful product answers as content, without requiring an FAQ block to pass the audit.

The most common finding is technically valid schema with semantically empty fields. A product description that reads “premium quality supplement for dogs” in the schema is technically correct markup, but it gives an AI agent nothing to differentiate your product from any other “premium quality supplement for dogs” in its index.

Source: Google Search Console: Rich Results Test

Fixing It on Shopify

Inspect your actual Shopify theme before editing. Shopify’s documentation supports JSON product templates whose HTML and Liquid live in referenced sections. Structured data may be produced by that Liquid code, a snippet or an app; there is no universal product.liquid location. Search for application/ld+json and structured_data, trace the output, and change the implementation responsible for it.

Use product metafields for approved facts that are missing from the product model, and render those facts visibly where they answer purchase questions. Map them into JSON-LD only through properties supported by the chosen destination. Shopify’s structured_data filter can generate product markup, but custom metafields are not automatically supported fields in every destination. Validate the generated output for representative products and variants.

Add a product-question section with answers verified against your own product records. Support tickets and reviews can help identify questions, but a competitor’s answer is not evidence for your product. Keep the answers readable; add FAQPage markup only for a documented consumer that uses it, not as a universal AI requirement.

For a headless frontend, generate supported structured data in the rendering layer from the same product records used for visible content. Check required fields, variants and missing values in the rendered output rather than assuming that a shared data model guarantees complete markup.

Source: Shopify: Product templates

Source: Shopify: The structured_data Liquid filter

What to Do This Week

Pick your ten highest-revenue products. Run each through Google’s Rich Results Test for the feature you intend to support, then compare the rendered markup with visible facts and the product record. Treat validation errors and factual discrepancies as separate findings.

For each product, identify recurring customer questions from support tickets and reviews. Publish verified answers where they help the purchase decision; there is no required number of FAQ entries or universal FAQPage block.

Fill missing required properties for your chosen product feature, plus supported optional properties backed by real data. Keep ingredients, certifications and use instructions accurate and visible even when the destination has no corresponding supported markup field.

Test the markup again after changes. Separately record dated prompts and answers from the platforms you want to observe, checking any cited details against your product record. A missing recommendation does not prove that a competitor has more complete structured data.