Structured data and schema.org: how to speak the language of Google and AI

Structured data is markup added to the code of a web page, most often in JSON-LD format using the schema.org vocabulary, to describe its content to search engines and AI systems.

In 2026, 48% of Google queries trigger an AI Overview: without clear semantic markup, a page becomes much harder to cite in these generated answers.

NEXUS SYNERGY, a WordPress and SEO agency with more than 150 sites delivered, details the schema types that are actually useful and the method to implement them without errors.

Structured data is semantic markup added to the code of a web page to describe its content in a language machines can interpret without ambiguity. It relies on the schema.org vocabulary and is most often written in JSON-LD, the format Google officially recommends.

This markup determines access to rich results and, increasingly, a page's ability to be understood and then cited by AI engines. At NEXUS SYNERGY, a WordPress agency specialized in SEO and AI visibility, we implement structured data on every site we deliver, and we have delivered more than 150 of them. This guide follows our method: definition, concrete role, format choice, priority types, implementation and validation.

What is structured data? Definition and double meaning

In SEO, structured data refers to markup, invisible to the visitor, that translates the content of a page into identifiable entities: an article, a product, a company, a question followed by its answer. In other words, you no longer describe your page with words the search engine has to guess at, but with properties it reads directly.

One clarification is needed, because the term is ambiguous. In computing, structured data refers to data stored in a predefined format, typically a SQL database, as opposed to unstructured data such as videos or emails. That is not our subject. This article covers the semantic markup of web pages, the kind that matters for your visibility on Google and in AI systems.

This markup relies on a common standard: schema.org, a vocabulary launched in 2011 by Google, Microsoft and Yahoo, later joined by Yandex. The project is huge and yet simple to use:

• Types: around 800 entities described, from Article to LocalBusiness, via Recipe or Event
• Properties: each type has its attributes, for example price and availability for a Product
• Syntaxes: JSON-LD, microdata or RDFa, three ways of writing the same vocabulary in code
• Open governance: the vocabulary evolves continuously, under an open license, with the search engines participating

In practice, about ten types cover 95% of a business site's needs. There is no point memorizing the whole vocabulary: what matters is choosing the right types and filling them in correctly.

What is structured data for, for Google and AI engines?

Structured data allows Google to display rich results: review stars, prices, breadcrumbs, recipe images or dated events directly on the results page. An eligible page occupies more visual space and attracts more clicks. According to Google's documentation, Rotten Tomatoes for example measured a 25% higher click-through rate on its marked-up pages.

The second role has taken on a new scale. Generative engines, AI Overviews first among them, need to identify reliable facts before restating them. And 48% of Google queries trigger an AI Overview in 2026. Marked-up content exposes its entities, its figures and its answers in a format these systems ingest without interpretation effort. It is one of the technical pillars of the GEO and AEO work we do for our clients, alongside the editorial strategy described in our guide on how to get cited by AI.

Concretely, clean markup acts on four levels:

• Eligibility for rich results: without valid markup, no chance of getting stars, prices or breadcrumbs
• Disambiguation: the engine distinguishes your brand, your authors and your products from entities with the same name
• Citability by AI: ChatGPT, Perplexity or Gemini connect an answer to its source more easily
• Knowledge Graph: the Organization properties feed your company's panel on Google

Let's be clear on one point: markup does not directly improve your ranking. It improves the understanding of the page and the presentation of the result, and therefore the click. The nuance matters, and we come back to it below.

JSON-LD, microdata or RDFa: which structured data format should you choose?

Three syntaxes let you write structured data, but the match was settled long ago. JSON-LD is written in a script block independent of the visible HTML, whereas microdata and RDFa scatter attributes across the page's tags. Google explicitly recommends JSON-LD, and our rebuild experience points the same way: markup separated from the template survives design changes, markup scattered through the HTML breaks at the first redesign.

Reading this table leaves little room for debate. A few situations nevertheless justify keeping another format:

• Legacy site in microdata: if the existing markup is valid, an immediate migration is not a priority
• Platform constraints: some older themes or modules inject their own microdata attributes
• Academic ecosystems: RDFa survives in semantic web projects, far from everyday SEO concerns

For any new project, our position is firm: JSON-LD, no exceptions. A developer reads it at a glance, a plugin generates it automatically and Google parses it with no risk of conflict with the HTML.

Which schema.org types should you prioritize for your site?

The right markup depends on the site's model, not on a universal list. Marking up everything schema.org offers would be counterproductive: you would add code with no benefit, because Google only uses about thirty types for its rich results. Here are the priorities we apply at NEXUS SYNERGY:

• All sites: Organization or LocalBusiness on the homepage, WebSite, BreadcrumbList on every page
• Brochure or corporate site: Service for each offering, Person for team and author profiles
• Blogs and media: Article or BlogPosting with author, publication and update dates, image
• E-commerce: Product with Offer, price, availability and AggregateRating when the reviews are real
• Local business: detailed LocalBusiness, opening hours, service area, geographic coordinates
• Question-and-answer content: FAQPage, useful for AI even though its rich result is restricted

A word of honesty on that last point. Google restricted the rich display of FAQs in 2023 to government and medical sites, and removed the one for HowTo. Should you abandon these markups? No. They no longer trigger a special display, but they keep structuring your answers for generative engines, which are not bound by Google's display rules.

The classic trap is stacking types with no coherence. One page = one main entity. The rest of the markup plays a supporting role, connected by identifiers, never competing.

How do you implement structured data on your site?

Implementing structured data always follows the same sequence, whatever the CMS: inventory the page templates, assign a schema.org type to each one, generate the JSON-LD, then validate. On WordPress, which powers 43.5% of websites according to W3Techs (2026), most of the work can be industrialized.

Our four-step process:

• Map: list the templates (home, services, articles, product pages) and the main entity of each one
• Generate: enable the schema graph of a plugin like Yoast SEO or Rank Math, which produces Organization, WebSite, Article and breadcrumbs automatically
• Complete: manually add the types plugins ignore, for example Service, FAQPage or an enriched LocalBusiness, via a code block or the plugin's fields
• Align: check that every marked-up property matches content visible on the page, without exception

Watch out for the default settings: a poorly configured plugin declares a generic Organization with no logo and no social profiles, and that is a missed opportunity. Take twenty minutes to fill in the company's complete identity.

At NEXUS SYNERGY, we find that about half of the sites we take over for a rebuild arrive with markup that is missing, duplicated or invalid. Fixing it is part of the technical foundation of our SEO services, on the same level as internal linking or performance. That foundation pays off: our clients see an average of +320% organic traffic, and Maison Laurent reached +450% in 8 months on a full SEO program, markup included. On the WordPress sites we build, the schema graph is in place from day one.

How do you test and validate your schema.org markup?

Untested markup is broken markup, or markup that will be. Validation takes ten minutes and relies on three free tools, to be used in this order:

• Google's Rich Results Test: checks the page's eligibility for rich displays and flags missing or invalid properties
• Schema.org validator: checks the complete syntax of the markup, including types Google does not use
• Search Console: the Enhancements section tracks detected items, errors and warnings across the whole site, continuously

The official Google documentation on structured data specifies, type by type, the required and recommended properties. Keep it open during implementation, because the requirements change: a recommended field can become required from one update to the next.

Do not validate only on launch day. A plugin update, a theme change or a migration can invalidate markup that used to work. A monthly pass through Search Console is enough to catch regressions before they cost you rich results.

What are the limits of structured data and the mistakes to avoid?

Structured data is not a magic wand, and selling it as one would be dishonest. It is not a direct ranking factor: Google has said so many times. Mediocre content with perfect markup remains mediocre content. Markup amplifies a page that already deserves to rank, it compensates for nothing.

Another limit: eligibility does not guarantee display. Google decides case by case, depending on the query and the trust granted to the site. You can therefore do everything right and wait several weeks before seeing the first rich results.

The mistakes we encounter most often in audits:

• Marking up invisible content: declaring reviews or an FAQ absent from the page violates Google's spam policies and exposes you to a manual action
• Self-assigned review ratings: an AggregateRating without real, verifiable reviews is the riskiest shortcut in technical SEO
• Duplicate markup: a theme and a plugin each declaring their own Organization create contradictory signals
• Empty or made-up properties: a JSON-LD filled with default values does more harm than no markup at all
• Markup never updated: outdated prices, hours or dates end up eroding the engine's trust

The rule that sums it all up: the markup must faithfully describe what the visitor sees. When you hesitate over a property, ask yourself whether a human would find the information on the page. If the answer is no, do not mark it up.

Frequently asked questions about structured data

Is structured data a ranking factor for Google?

No, structured data is not a direct ranking factor. Google uses it to understand the content of a page and to decide on rich displays, not to push it up the rankings. The effect on traffic exists nonetheless, but it takes another path: a rich result occupies more space and attracts more clicks at equal position, as shown by the Rotten Tomatoes case with 25% more clicks on marked-up pages. Markup also improves the understanding of the site's entities, which makes your whole SEO more reliable. An indirect lever, then, but a very real one.

What is the difference between structured and unstructured data?

The difference lies in the context in which the term is used. In data management, structured data follows a predefined format, typically SQL tables, whereas unstructured data covers free text, images, audio and video. In SEO, the expression means something else: the schema.org markup added to the code of a web page to describe its content to search engines. The two notions share a common idea, giving machines a readable format, but they involve neither the same tools nor the same professions. This article deals exclusively with web markup, the kind that influences your visibility on Google and in AI systems.

Is the JSON-LD format mandatory for structured data?

No, JSON-LD is not mandatory, but it is the format Google officially recommends and the one we advise without reservation. Microdata and RDFa remain accepted and correctly interpreted by search engines. The difference plays out in maintenance: a JSON-LD block lives separately from the visible HTML, so it survives visual redesigns and is easy to validate. Microdata attributes, scattered through the page's tags, break as soon as an integrator touches the template without realizing it. If your site already uses valid microdata markup, there is no urgency to migrate everything. For any new implementation, however, go with JSON-LD.

Do you need a plugin to add structured data on WordPress?

No, a plugin is not essential, but it saves considerable time on standard markup. Yoast SEO and Rank Math automatically generate a complete graph: Organization, WebSite, Article and breadcrumbs, connected to each other by clean identifiers. For a typical blog or brochure site, this foundation covers most of the need. More specific types, such as Service, Event or a detailed LocalBusiness, often require a manual JSON-LD addition in the relevant template. The real issue is not the tool but the configuration: a plugin left on its default settings produces generic markup, with no logo and no complete identity.

Does structured data help you get cited by ChatGPT and Perplexity?

Yes, structured data is among the signals that make citation by generative engines easier. These systems need to extract reliable facts before rephrasing them: marked-up content exposes its entities, its figures and its answers in a directly usable format, which reduces the risk of being misinterpreted or ignored. Markup is not enough on its own, because citability depends first on the quality and clarity of the content, the site's authority and its structure. That is the purpose of GEO, the discipline that optimizes visibility in generative answers, and schema.org markup is its technical foundation.

How long does it take for rich results to appear?

Expect anywhere from a few days to several weeks after putting valid markup in place. Google first has to recrawl the pages concerned, then decide to activate the rich display, and that decision remains at its discretion: technical eligibility never guarantees display. Three ways to speed things up: request inspection of your strategic URLs in Search Console, check that the markup passes the Rich Results Test without errors, and monitor the Enhancements section to fix warnings quickly. If nothing appears after two months despite clean markup, the problem usually lies elsewhere: insufficient site trust, content that is too weak or queries that do not trigger this type of display.

Structured data belongs to those quiet technical projects that separate a properly built site from a genuinely visible one, on Google as in AI answers. Want to know what your site's markup is worth? NEXUS SYNERGY offers a free initial audit in three parts, technical, SEO and AI visibility, with a firm quote within 24 hours: contact our team to talk about it.