Structured data: what schema actually buys you
Structured data is a machine-readable description of what a page contains, added alongside the visible content. Google defines it as a standardized format for providing information about a page and classifying the page content. The vocabulary comes from schema.org, and the format Google recommends is JSON-LD: a small script block containing a JSON object that states, in plain fields, that this page is an article with this headline, author and date, or a product with this price and availability. Because it sits in its own block rather than being woven into the HTML, JSON-LD is the easiest format to generate, template and maintain at scale.
The rich results schema powers
Structured data earns its keep in classic Google Search, where it makes pages eligible for rich results: enhanced listings that show more than a title and a snippet. According to Google's structured data documentation, the point of the markup is exactly these features:
- Product snippets. Price, availability and review ratings shown directly in the result. For shops this is the most valuable rich result, and it pairs with the Merchant Center feeds covered in local and merchant data.
- Article markup. Headline, author and date information that helps Google present news and blog content correctly across Search surfaces.
- Recipes, events, job postings and breadcrumbs. Each has its own markup type, its own required fields and its own eligibility rules.
One cautionary tale belongs here. FAQ rich results, once a popular optimization target, were restricted in August 2023 to well-known, authoritative government and health websites, and in 2026 Google retired the feature from Search entirely. The lesson is general: markup buys eligibility for a feature, never a guarantee, and features can be withdrawn. Check the current documentation before investing in any type.
Validation
Markup only counts if it parses. Google's Rich Results Test checks whether a page's structured data makes it eligible for Google features and can preview some of them. The Schema Markup Validator checks general schema.org syntax without the Google-specific layer. After deployment, the rich result status reports in Search Console show whether pages keep their eligibility. One rule sits above all tooling: the markup must describe content that is actually visible on the page. Markup that promises what the page does not show is treated as spam.
The honest limit: AI search does not require it
Google is unambiguous on this point. Its AI features guide states that structured data is not required for generative AI search and that there is no special schema.org markup you need to add. AI Overviews and AI Mode select sources through retrieval grounded in the normal ranking systems, and the actual entry requirement is that a page is indexed and snippet eligible, which is covered in snippet eligibility. Anyone selling schema as a direct lever on AI answers is contradicting Google's own documentation.
When to still invest
Schema remains worth doing where the classic features still pay. If you sell products, product markup plus a Merchant Center feed puts price and availability into results. If you publish articles, article markup is cheap at publish time and keeps authorship and dates machine-readable. Vupie includes article schema on every page it publishes for exactly this reason. What to skip is equally clear: retired features like FAQ, and any markup describing content the page does not visibly contain. Treat schema as inexpensive insurance for rich results, not as an AEO shortcut.