Schema.org and structured data for online shops: making your catalog machine-readable

Why structured data still matters for AI-era visibility

Say a shopper asks ChatGPT: "Where can I find a waterproof hiking jacket under 150 euros in size L?" It isn't scanning product photos or beautifully designed landing pages to answer that. It's looking for machine-readable facts: product type, price, availability, size, material. Those are exactly the facts Schema.org is built to carry. Without it, your shop is invisible to that kind of query even if a human could find the jacket in ten seconds — the facts are there, just not in a form the system can extract.

That's a real shift from classic SEO . A good page and a few rich snippets used to be enough for a Google ranking. AI systems like Perplexity or Google AI Overviews assemble an answer instead of listing links, so they need structure to pull from. Cleaner, more precise data makes it easier for these systems to cite your product with a price and a link attached. It's not a guaranteed ranking boost — Google is explicit that no special markup is required for AI Overviews — but it is the difference between being extractable and being noise.

For you as a shop operator, the takeaway is this: structured data isn't a technical footnote to hand off and forget — it's table stakes for being legible to any system parsing your site. Skip it, and you don't just lose ranking positions, you risk not being parseable at all.

Product, Offer, and AggregateRating: the three schema types worth getting right

For an online shop, the Product type is the foundation. It describes a single item: name, brand, description, GTIN or SKU, image. That sounds basic, but it's routinely done poorly — many shops fill in the name and description and skip GTIN, brand, and material. Those detail fields are exactly what lets a crawler or an AI tell your product apart from ten near-identical listings.

The second type is Offer, nested inside Product: price, currency, availability (InStock, OutOfStock), and ideally a price-validity window. If your jacket is on sale at 129 euros and in stock, the markup needs to say exactly that. A system filtering for "under 150 euros and available" can only surface you if those values are accurate and current.

The third piece is AggregateRating and Review. Ratings matter because they give a system a trust signal to point to — a product with a real, product-specific rating reads as more credible than one with none at all. But be careful: Google penalizes invented ratings or a single site-wide score applied across every product. Only mark up what's real and specific to that product.

JSON-LD vs. Microdata: the format that's easier to keep correct

There are three ways to embed Schema.org markup: Microdata, RDFa, and JSON-LD. For e-commerce, JSON-LD is the clear choice. It sits as a single data block, usually in the head or at the end of the page, separate from the visible HTML. That makes it easier for crawlers and parsers to read cleanly, and easier for you to maintain, since it lives in one place instead of being scattered through the template.

Microdata nests attributes directly into your product page's HTML. In a template that's evolved over years, that breaks or goes half-finished the moment someone touches the design. JSON-LD separates content from presentation: when your frontend team redesigns the product tile, the structured data block underneath doesn't have to change at all. That separation is what actually keeps data quality intact day to day.

In practice: Shopify, Shopware, and WooCommerce themes often generate JSON-LD automatically, but rarely completely. Check the source of an actual product page for price, availability, and GTIN. More often than not, the exact field a filtering query needs is the one that's missing.

The mistakes that make your shop's data unreliable

The most common mistake is contradictory data: the JSON-LD says one price, the visible page shows another, because a discount plugin updates the frontend but never touches the markup. Any system that catches the mismatch has to treat your data as unreliable. Keeping the visible price and the markup in sync isn't optional polish — it's the baseline requirement.

Second: stale availability. Your bestseller sells out, but the markup still says InStock. If an AI assistant recommends it anyway and the customer lands on an unavailable product, that costs you twice — the sale, and your credibility as a reliable source going forward. Availability needs to flow automatically from your inventory system into the markup, not get updated by hand.

Third: unclear variants. A shirt in five colors and four sizes needs each variant marked up properly via ProductGroup and hasVariant. Cram every variant into a single Product and there's no way for a system to isolate "blue in size M." In fashion and furniture e-commerce especially, this alone can make you invisible to specific queries.

Beyond the product page: FAQ, Organization, and shipping details

Structured data doesn't stop at the product. FAQPage markup is genuinely useful for shops because AI systems assemble answers out of concrete question-and-answer pairs. If your product page answers "Does this run large?" or "How long is delivery?" in structured form, you're handing the assistant a ready-made building block for its answer — which improves the odds your shop gets cited as the source.

Just as important is Organization markup for your brand: company name, logo, contact details, social profiles. It helps a system recognize your shop as a real, identifiable entity rather than an anonymous domain. Smaller shops especially tend to underestimate how much a clean entity record adds to perceived legitimacy.

Newer extensions like shippingDetails and hasMerchantReturnPolicy on the Offer are increasingly relevant for e-commerce. Shipping cost and return policy are often the facts that decide a purchase. Mark them up and you give an assistant something concrete to fold into its answer — an edge over competitors who've only bothered with price and title.

How to check whether your markup actually works

Never ship structured data without validating it. The Schema Markup Validator and Google's Rich Results Test will both tell you immediately whether your JSON-LD is syntactically correct and complete. Run a real product URL through both and confirm that price, availability, and rating are actually recognized. Catch a missing field here, before it costs you somewhere you can't see.

The second, often-skipped test is the practical one: ask the AI itself. Put a realistic customer question about your assortment to ChatGPT, Perplexity, or Google AI Overviews and see whether your shop turns up, and whether the price and details it gives are correct. This reality check tells you more about your actual AI visibility than any validator can, because it tests the whole chain through to the actual answer.

Do this on a schedule, not once. Prices, assortment, and availability shift constantly. A lightweight monitoring check that spot-checks markup against the live frontend keeps you from sitting in AI answers for weeks with wrong data and not knowing it.

Where to start first when time is limited

You don't have to do everything at once. Start with your highest-revenue products or categories and give them complete Product and Offer markup. A small slice of your catalog usually accounts for most of the demand that's worth optimizing for. One thoroughly marked-up bestseller beats a hundred half-tagged slow movers.

Next, lock down data quality: automate the flow of price and availability from your shop system into the markup so contradictions can't creep in. Only after that do additions like FAQ, shipping, and return data start to pay off. That order — correct first, complete second, enriched third — keeps you from investing effort in extras while the fundamentals are still shaky.

Remember: structured data is an ongoing practice, not a one-off project. A shop that keeps its data consistent and current gets treated as a reliable source and cited repeatedly over time. That consistency, not a single implementation sprint, is what actually compounds.

A 30-day plan: from audit to complete markup

Don't start with your entire catalog — start with your 20 highest-revenue products. In week one, check what schema your shop software already outputs automatically. Shopify, Shopware, and WooCommerce all ship a basic Product markup out of the box, but it's almost always incomplete — price, availability, or the return window is missing. Note, per template, which fields come up empty.

In weeks two and three, fill the gaps, either through a plugin or directly in the template. Build one clean template per product type rather than patching products one by one. In week four, roll the same logic out to category and brand pages. By the end of the month you have a system that generates correct markup automatically for every new item, without you touching it again.

GTIN, MPN, and brand: the identifiers that make your product unambiguous

AI systems and shopping feeds match products by unique identifiers. Enter the GTIN (the number behind the barcode), the MPN (manufacturer part number), and the brand in your Product markup, and a system can match your listing to the same product carried by other shops. That match is what decides whether you even show up in a price or recommendation comparison at all.

For private-label products without a GTIN, use MPN and brand instead — don't just leave the field blank. Make sure those values match exactly what's in your Google Merchant feed. If the feed and the markup disagree, a system has to discount both signals, because it can't tell which one to trust.

A practical test: search your GTIN on a price-comparison site. If your product shows up correctly, your data foundation is solid. If it doesn't appear at all, or shows up under the wrong name, you have a matching problem that no amount of polish on the product page will fix.

What markup can't fix

Structured data amplifies what's already there — it doesn't manufacture substance out of nothing. If your product copy is thin, your delivery times are vague, or you have no reviews, the cleanest JSON-LD in the world won't change that. Markup helps a system understand your offer faster; it can't substitute for the offer being good.

Don't expect an immediate effect, either. It can take weeks for AI systems to re-crawl your updated markup and start factoring it into answers. Measure the impact over months, not days — for example by tracking sessions landing on your product pages from AI assistants and generative search referrals.

And one clear warning: marking up reviews or availability that don't actually exist gets caught. False AggregateRating data can lead to manual penalties and the loss of rich results entirely. Keep your markup honest and in sync with what a customer actually sees on the page.