Structured data for ecommerce: What must product pages get right?

Structured data labels information so search engines can interpret it more reliably. Product describes the item; Offer describes the associated commercial offer. A common format is JSON-LD: a block of data within the HTML, rather than extra copy displayed to visitors.

For pages where customers can buy a product, Google documents merchant listings. Accurate information can establish eligibility for enhanced product displays, such as price and availability. It does not force Google to show them. Passing a test is not a ranking promise either. Google's introduction to product structured data explains the distinction.

Our priority is therefore not the number of schema types installed. It is whether the information describes the actual item being sold and stays accurate when something changes.

We recommend a small but deliberate sample: a simple item, a discounted product, an unavailable item and a product family with variants. Check these six points for each URL:

• Product identity: Do the name, brand and available product identifiers match the item being offered?
• Price and currency: Does the published amount match the visible variant and target market?
• Availability: Does the machine-readable value reflect the actual ordering status?
• Variants: Do the image, identifier and offer belong to the correct size, colour or specification?
• Reviews: Do marked-up reviews actually relate to this product, and can visitors find them on the page?
• Updates: Are the values still correct after a price import, stock change or promotion ending?

Not every property is required for every search feature. Technical implementation must follow the relevant Google merchant listing requirements. A missing GTIN, for example, cannot be fixed by inventing a number. Our list is a practical merchant checklist, not a complete technical specification.

Consider a simplified testing example: a retailer sells insulated bottles in two sizes. The smaller bottle costs €24.90 and the larger one €29.90. Selecting the larger bottle updates the visible price correctly. The structured offer still contains the identifier and price of the smaller version. The selector appears to work, but the data mapping is wrong.

Google supports ProductGroup to describe related variants. The implementation also depends on whether customers select variants on one shared page or visit separate URLs. Google's product variant documentation explicitly distinguishes these approaches.

For acceptance testing in this example, we would open both sizes individually, check their destination URLs and compare the visible offer with the extracted data and any product feed. Testing only the default selection is not enough. This illustrates a testing process; it is not a measured client case study.

Not necessarily. The official PrestaShop Classic theme already includes a template for product data in JSON-LD. You can inspect it in the Classic theme source code. That does not mean every installed theme and store version produces the same output.

As a PrestaShop agency, we first examine the existing store: what does the theme generate, what does a module add, and where do price and stock information come from? Two outputs are not automatically wrong simply because there are two. The problem arises when they describe the same item inconsistently. There should then be a clear responsibility for generating the data.

PrestaShop provides a useful foundation when product data and its presentation need to reflect a retailer's particular catalogue. We would use that flexibility deliberately instead of installing another module as a precaution. An initial assessment establishes whether the work involves theme changes, module configuration or custom development. The same testing principle applies to Shopify and Shopware.

A correct value in the ERP does not prove that every downstream system has received it. Imports, store data, caches, HTML output and feeds can update at different times. We therefore recommend tracing one specific item through the whole process and recording where an incorrect value first appears.

Structured data does not replace a product feed. Google describes website markup and submission through Merchant Center as complementary ways to supply information. Consistent data helps with matching and verification. Source: sharing product data with Google.

For a retailer offering prices in euros and Swiss francs, testing must also cover the target market. A German-language page alone does not establish which currency or offer applies to that URL. These rules need to be agreed alongside the source data and store configuration.

It helps describe existing information clearly. That does not establish a preferential position in an AI answer. For Google AI Overviews and AI Mode, Google does not require a special AI schema or an additional AI file. Established SEO fundamentals remain relevant. Source: Google's guidance on AI features and websites.

For merchants, our approach is to establish reliable product information and an accessible website first. We can then assess how the store appears in search results and AI answers. A guaranteed ChatGPT recommendation based solely on schema markup would go beyond what the evidence supports.

We would start by checking the selected pages with Google's Rich Results Test and comparing the output with the visible product page. A syntactically valid price can still be the wrong price. The general structured data guidelines require markup to represent the actual page content appropriately.

After a correction, acceptance testing should include another check following price and stock changes. For ongoing monitoring, Google Search Console helps identify affected page types and record issues for investigation after updates. Technical validity, actual search appearance and orders are different measures and should be assessed separately.

At sjmedia, we connect ecommerce SEO with technical PrestaShop support. A typical product page and a problematic variant give us a concrete starting point: is the issue in the source data, the transfer or the final output? Answering that question turns a vague SEO concern into a defined implementation task.