Descriptive Schema vs. Cosmetic Schema

When 100% Schema.org validation doesn’t mean what you think it does

A brief technical guide for IT and Marketing teams: how to determine in ten minutes whether your structured data is merely cosmetic or truly descriptive

If you have run your corporate website through Google’s Rich Results Test and the report came back green, it is tempting to consider the matter closed. The validator says everything is fine, so everything must be fine.

Not necessarily.

There is a real difference between structured data that simply validates and structured data that actually serves a purpose. Most WordPress implementations—especially those managing multiple brands or product lines—unknowingly fall into the former category. And the problem isn’t detected by the validator; it comes to light when an AI system tries to understand what the company sells and finds nothing to extract.

Two deployments, same seal of approval

A typical example. A product catalog page on WordPress passes Google validation 100%. The JSON-LD graph includes:

• A CollectionPage with a name and URL
• A complete BreadcrumbList with all positions
• A WebSite with a search action
• An Organization with a logo

Everything is correct. Zero errors, zero warnings. And yet, that graph doesn’t say which products are in that collection, nor does it list their characteristics or how the brand relates to its parent company. It is like a library catalog card inside a book cover with no book inside.

That is what we call a cosmetic implementation: it meets the minimum requirements of the SEO plugin to avoid errors, but it fails to describe the business. This differs from a descriptive implementation, where the graph lays out the actual entities—products, technical specifications, corporate relationships—in such a way that any system, whether human or machine, can reconstruct what the page is about without needing to read the entire HTML.

This difference doesn’t show up in any validation report because the validator only checks for correct syntax, not for the completeness of the information.

Why this is also a matter for Marketing, not just IT

“Cosmetic” structured data goes unnoticed not only by Google but also by the generative AI systems that now answer questions about your company before a customer even visits your website. If your data graph does not accurately describe what each business line sells, ChatGPT, Gemini, or Copilot (and others) have no reliable source from which to extract that information—and when they cannot find it, they infer it.

Sometimes they get it right. Sometimes they invent a product you don’t offer, or describe your company as the sector’s most visible competitor because the context didn’t distinguish between the two.

That reputational risk is not hypothetical: it is the direct and silent consequence of a cosmetic rollout.

Why this is no longer just about SEO

For years, this was—at worst—a missed opportunity to appear with enhanced features in search results. Today, the situation is different. Generative AI systems (ChatGPT, Copilot, Gemini, Google’s generated answers) construct their responses based on data they can reliably extract. If a page’s data graph lacks substance, these systems either ignore the content or interpret it based on unstructured text, carrying the inherent risk of error or omission.

For an IT department, this has a concrete interpretation: it’s a data governance problem, not a marketing one. The SEO plugin managing the website is publishing structured statements about the company (name, corporate hierarchy, catalog), and no one is likely verifying whether those statements are complete or correct.

Five questions you can answer in ten minutes

1. Open any category page or product listing on your website. Look for “View source code” and locate the application/ld+json block. Does it describe the products on that page, or only the navigation (breadcrumbs, search bar, menu)?
2. If the company operates multiple brands or divisions under a single domain or multisite setup, does the JSON-LD for each brand link it to the parent organization, or does each brand appear as an isolated entity with no declared relationship?
3. Look for the description field within the WebSite or Organization blocks. Is it populated with an actual sentence, or is it empty?
4. Run the URL through Google’s Rich Results Test. Does the report show items as “valid with optional warnings”? That warning—often overlooked—usually points not to the data needed to move from “cosmetic” to “descriptive,” but rather to the data Google requires for its own needs.
5. Ask a generative AI which specific products a particular business line sells, providing only the page URL. If the answer is vague, generic, or completely made up, the problem isn’t the AI; it’s that there is no structured data for it to read.

If two or more of these answers have left you with more questions than certainties, the current implementation—if one exists—is likely merely cosmetic.

What sets a descriptive implementation apart

It is not about adding fields just for the sake of it. It is about ensuring the data graph reflects real business entities: products with their technical attributes, brands explicitly connected to their parent organization, and catalog pages that list their contents rather than simply describing the navigation. It is the difference between a map that only marks streets and one that also indicates what is inside each building.

This is not limited to a single sector. It appears just as frequently among industrial manufacturers, B2B distributors, and groups with multi-site or multi-brand architectures—precisely because these are the scenarios involving the most interconnected entities, and where standard SEO plugins offer the least help.

If this sounds familiar

You are not alone; this is the default pattern for almost any WordPress site using a standard SEO plugin without explicit semantic governance. Fixing it does not require rebuilding the website, but it does require a review led by someone who can distinguish between cosmetic and descriptive data—and who knows what to prioritize based on the business model.

This isn’t just for WordPress.

This popular CMS is often the focus of my work, but I do not limit myself to WordPress. Any business website can undergo this evolution; it is implemented in parallel with the content, requiring no aesthetic changes or editing of existing content.

---

Ricard Menor has over 20 years of experience in SEO consulting, specializing in structured data architecture and semantic governance for complex web projects, including multi-site and multi-brand WordPress deployments in the industrial sector. His methodology is publicly documented under the DSH (Homogeneous Semantic Descriptor®) framework.

More information: https://www.seofreelance.eu/freelance-seo-consultant/