4/11/2026
Structured data is machine-readable markup added to a page to describe what it contains: an organisation, a product, an article and its author, an event, a question and its answer. Its reference vocabulary, schema.org, is an open, community-maintained vocabulary: it does not belong to Google. This guide covers the method, the formats and the prioritisation: what to mark up, in which format, how to check it, and what traditional search engines and generative engines actually do with that markup.
Structured Data: Definition, SEO Benefits and GEO Impact (AI Search Engines)
Structured data does not replace HTML; it complements it by translating a page's key information (product, price, author, organisation, event) into a strict language that crawlers can interpret without ambiguity.
Schema.org was launched in 2011 by Google, Microsoft and Yahoo, later joined by Yandex, and it is maintained by a community linked to the W3C. What Google displays is therefore only one use of the markup, not its reason for existing.
Why Structuring Your Information Changes Understanding and Display in the SERP
To structure information is to represent it using a predefined format: a type, properties and expected values (dates, amounts, URLs, identifiers). A person immediately understands that an amount next to an item name is a price; a crawler must infer that meaning from the layout, CSS classes or surrounding text.
With a standardised vocabulary, you make three levels of information explicit:
- the page's main entity: a product, a service, an article, an organisation;
- its attributes: price, availability, reviews, author, date, address, coverage area;
- relationships between entities: organisation and brand, product and offer, article and author.
Who Reads Structured Data: Search Engines, Social Networks, Assistants and AI Crawlers
The same markup serves many systems:
- search engines (Google, Bing, Yandex, Apple): understanding the page and eligibility for rich results;
- social networks and messaging apps: link previews with title, image and description;
- voice assistants and aggregators: extraction of precise facts (opening hours, address, price, date);
- generative engine crawlers (GPTBot and OAI-SearchBot for OpenAI, ClaudeBot, PerplexityBot, Googlebot for Gemini, AI Overviews and AI Mode): they read the HTML, and therefore the markup it contains.
This last use is gaining weight: 39% of French people use AI engines for their searches (IPSOS, 2026). Accurate markup describes your business in the same way to all of these readers.
What Structured Data Changes for SEO
For Google, markup makes understanding easier (entity, attributes and relationships are stated rather than inferred) and makes the page eligible for certain rich results: a product's price, an article's image and date, an organisation's logo.
Structured data is not a direct ranking factor: a thin page, or one poorly aligned with search intent, will not climb thanks to its JSON-LD. The gain lies in a more informative snippet and facts that are reproduced without error.
Two limits remain structural. Eligible does not mean displayed: Google decides based on the query, the device and the context, even for valid markup. And markup must match the visible content; otherwise the page loses its eligibility and is exposed to a manual action for spammy structured data.
What Google Still Displays, and What It No Longer Shows
Several rich results have disappeared or been restricted since 2023, even though many guides still recommend them. A type with no Google display can nevertheless keep its structuring value for other systems.
High-Impact Schemas: Organization, WebSite, BreadcrumbList, Article, Product, LocalBusiness
The following table reviews, type by type, how each one is displayed in Google at the time of writing and what the markup provides outside Google:
| schema.org type | Google display (at the time of writing) | Value beyond Google display |
|---|---|---|
| Organization | Active: logo and organisation information | Identifies the company, acts as the pivot node (@id, sameAs) |
| Product + Offer | Active: price, availability, compliant reviews | Exact product and offer attributes |
| Article / BlogPosting | Active: headline, image, date, author | Attribution, author, publication date |
| LocalBusiness | Active: address, opening hours, contact details | Reliable contact points |
| Event, VideoObject, SoftwareApplication, JobPosting, Recipe, Dataset | Active, subject to each type's eligibility criteria | Dated facts and standardised attributes |
| BreadcrumbList | Shown on desktop; breadcrumbs no longer shown on mobile since early 2025 | Explicit site hierarchy |
| Review / AggregateRating | Active on Product, SoftwareApplication and eligible types; reviews of your own organisation are not eligible | Summary of the reviews actually displayed |
| FAQPage | FAQ rich results restricted since August 2023 to authoritative government and health websites; no display for a B2B site | Explicit question / answer structure |
| HowTo | HowTo rich results removed in September 2023; the type remains valid in schema.org | Explicit structure of the steps |
| WebSite + SearchAction | Sitelinks search box removed at the end of 2024 | WebSite keeps its role: site name, link to Organization |
For pagination, Google has not used rel=prev/next since 2019. For FAQPage and HowTo, two things are true at once: these types explicitly structure questions, answers and steps for the systems that read them, AI crawlers included, and no quantified evidence shows any effect of this markup on their answers. To decide case by case, see how to mark up an FAQ with FAQPage since the 2023 restriction.
The context has also changed: more than 50% of Google searches show an AI Overview (Squid Impact, 2025), above the traditional rich results.
Structured Data Formats: JSON-LD, Microdata and RDFa
Three formats express the same schema.org vocabulary, and Google reads all three. They differ in syntax, placement and maintenance cost:
| Format | Syntax | Placement | Maintenance | Google's position |
|---|---|---|---|---|
| JSON-LD | <script type="application/ld+json"> block | <head> or <body>; JavaScript injection accepted | Separate from the HTML, easy to generate per template | Recommended format |
| Microdata | itemscope, itemtype, itemprop, itemid, itemref attributes | On visible HTML elements | Tied to components: to be reviewed at every front-end redesign | Read and supported |
| RDFa | vocab, typeof, property attributes | On visible HTML elements | Tied to components, more common in semantic web projects | Read and supported |
JSON-LD: The Most Robust and Easiest Format to Maintain at Scale
Google recommends JSON-LD because it separates the data layer from the display HTML. A front-end redesign, an A/B test or a component change does not break the markup. The block is generated from CMS fields, which lets you industrialise across thousands of pages with a single rule per template.
Microdata and RDFa: When They Remain Relevant
Microdata annotates the visible elements directly: the information stays tied to the displayed content, but every component change can break a property. Microdata remains relevant on a site that is already marked up this way and is stable. For attribute details and complete examples, see how to write Microdata markup and migrate to JSON-LD.
RDFa is mainly used where knowledge graphs are already managed. Whatever the format, keep a single source of truth per page: no duplication between a CMS plugin and the template.
Which Types for Which Page: E-Commerce, B2B and SaaS
Start with pages that combine business impact, stable demand and verifiable information, choose one main type per template, then keep the same entities and the same @id identifiers from one page to the next. For the code of each type, rely on our library of JSON-LD examples by schema type.
E-Commerce Markup Plan: Homepage, Categories, Product Pages, Reviews, Local Pages, Blog
In e-commerce, markup enhances the product snippet (price, stock, reviews) and clarifies the catalogue (variants, offers, brand). Here is the markup plan by page type, without the features Google no longer uses:
| E-commerce page type | Schemas to prioritise | Key considerations |
|---|---|---|
| Homepage | Organization, WebSite | Official entity, logo, sameAs; no SearchAction (search box removed at the end of 2024) |
| Category page | BreadcrumbList, ItemList | Do not mark it up as a Product; ensure a clear hierarchy |
| Pagination | BreadcrumbList, ItemList | Consistent canonical; rel=prev/next is no longer used |
| Product detail page | Product, Offer, AggregateRating if reviews are displayed, BreadcrumbList | Exact, visible price, currency and stock; product identifiers; accessible images |
| Buying guides, advice | Article / BlogPosting, FAQPage if the questions are real | FAQPage has no Google display; only mark up what is visible |
| Local pages (stores, click-and-collect) | LocalBusiness, Organization, BreadcrumbList | Address and opening hours identical on the page and in other sources |
| Blog / magazine | BlogPosting, Person (author), BreadcrumbList | Date, author, image |
| Proof pages (comparisons, tests) | Article, VideoObject if video | Only mark up reviews if they genuinely exist and comply with the guidelines |
The most frequent pitfalls come from the catalogue itself:
- Variants: maintain a clear approach between parent product and offers, without conflicting fields.
- Stock and price: a mismatch between the page and the markup loses the enhancement and distorts the facts reused elsewhere.
- Facets: filtered pages that resemble categories multiply quickly and create inconsistencies.
- Reviews: only declare real, visible reviews with coherent aggregation.
B2B and SaaS Markup Plan: Organisation, Services, Software, Authors, Events
On a B2B or SaaS site, the aim is to make explicit who you are, what you offer and who is writing; a plain "WebPage" on a service page says nothing about the offer. The following table matches each page type with its schemas:
| B2B / SaaS page type | Schemas to prioritise | What the markup makes explicit |
|---|---|---|
| Homepage | Organization (pivot node), WebSite | Identity, logo, official profiles via sameAs |
| Service page | Service (provider pointing to Organization), BreadcrumbList | Nature of the offer, area served, provider |
| Software product page | SoftwareApplication, Offer | Application category, price, reviews if displayed |
| Local page (office, agency) | LocalBusiness, Organization | Reliable contact details and opening hours |
| Blog / resources | Article / BlogPosting, Person (author) | Author, expertise, date, publishing organisation |
| Webinar, dated training | Event | Date, venue or online format, registration |
| Study, barometer, dataset | Dataset | Scope, method, date, licence |
| FAQ | FAQPage | Understanding of questions / answers, no Google display |
| Case study, proof page | Article, VideoObject if video | Structured, attributed proof |
The costliest error is not defining the entity behind the content: without an author linked to a Person or a provider linked to the Organization, attribution is lost.
How to Ensure Your Data Is Structured: A Step-by-Step Method
The most robust approach treats markup as a machine output of your templates, in the same way as the meta description. It comes down to four steps:
- Step 1, inventory: list the entities your site actually exposes (organisation, products, services, locations, authors, events), then the highest-value templates.
- Step 2, choose the types: one main type per template (
Product,BlogPosting,Service…) and the required or recommended properties for that type. - Step 3, ensure compliance: markup must be a faithful extract of the visible content; nothing invented, no reviews that are not shown, no different price.
- Step 4, generate per template: map each CMS field to a schema.org property, validate on a sample of URLs, then roll out across the whole template.
Document and version your mapping rules (property, CMS field, expected format): a change to the data model should trigger a markup update followed by a new test. This work aligns with your GEO content strategy: the entities you mark up are the ones your content must name clearly.
Full Example: Turning a Page Section into JSON-LD
Consider a simple product page. Here is its display HTML, where the crawler must guess what constitutes the name, image, description, price and stock status:
<article><h1>Ergonomic office chair</h1><img src="/img/ergonomic-chair.jpg" alt="Ergonomic office chair"><p>Adjustable chair with lumbar support.</p><p>Price: €199.90</p><p>Availability: in stock</p></article>
The same content, described in JSON-LD, makes each field explicit with a type, a coded currency and a standardised availability value:
<script type="application/ld+json">{"@context": "https://schema.org","@type": "Product","name": "Ergonomic office chair","image": "https://www.example.com/img/ergonomic-chair.jpg","description": "Adjustable chair with lumbar support.","offers": {"@type": "Offer","price": "199.90","priceCurrency": "EUR","availability": "https://schema.org/InStock"}}</script>
Notice the rigour: a decimal price, an ISO currency code and an availability value taken from a closed list. That standardisation makes the information portable from one site, and one reader, to another.
Where to Place the JSON-LD Script, What to Include and What to Avoid
Three rules are enough for most templates:
- Where to place it: in the
<head>or the<body>; for Google, the position does not matter as long as it is in the rendered HTML of the relevant page. - What to include: key information that users can verify (name, description, image, price, stock, author, date, contact details).
- What to avoid: invented fields, reviews that are not shown, a price that differs from what is visible, an entity that does not match the page's main topic.
Structured Data and Generative Engines: What Markup Provides, and What It Does Not
At the time of writing, AI Overviews and AI Mode have been available in France since summer 2026, alongside ChatGPT, Perplexity, Gemini and Claude. Their crawlers read the HTML, and therefore the markup. The topic is covered in our guide to structured data and generative engines; here are the essentials.
What Markup Provides to Generative Engines
Structured data helps a system understand a page. It tells it, without ambiguity:
- which entity the page describes and what type of content it is;
- who the author is and which organisation publishes it;
- when it was published or updated;
- which facts it carries: price, address, availability, opening hours.
It therefore supports accurate reproduction (the right name, the right figure, the right attribution); stable @id values and sameAs links make consistency easier between your pages and other sources.
What It Does Not Do: No Proven Direct Effect on Citations
No study shows that adding structured data gets a page cited by an AI. The available figures concern editorial structure, not markup: pages with an H1-H2-H3 hierarchy are 2.8 times more likely to be cited by AI systems (State of AI Search, 2025). This is a correlation, not causation (see our GEO statistics).
For generative engines, markup is only one piece of a whole, described in our guide to Generative Engine Optimization (GEO):
- entities named explicitly in the text;
- structured content: a single H1, H2s and H3s, lists, tables;
- consistent information between the page, its markup and other sources;
- reachable pages, with no AI crawlers blocked by mistake in the robots.txt file.
Testing, Validating and Monitoring Structured Data
Three tools complement each other. The Rich Results Test checks eligibility for Google's rich results, on a URL or a code snippet. The Schema.org validator (validator.schema.org), which succeeded the Structured Data Testing Tool, retired in 2020, checks compliance with the vocabulary for all types. Google Search Console then monitors indexed pages: Enhancements reports by type and URL Inspection.
An error makes the page ineligible; a warning flags a missing recommended property. Fix errors first and re-test after every template change or migration. The detailed procedure is in our guide on how to test your structured data with the Rich Results Test.
Common Errors: Effects and Fixes
Markup errors stem more from templates and CMS data than from syntax. The following table summarises the most frequent ones:
| Error | Effect | Fix |
|---|---|---|
| Marking up information that is not visible on the page | Loss of eligibility; risk of a manual action for spammy structured data | Only mark up what is visible and accurate |
| Duplicates from a CMS plugin + template | Two contradictory entities on the same page | A single source of truth per page |
Unstable or missing @id values | Entities not linked from one page to another | Fixed @id values and a single Organization node |
| Reviews of your own organisation | Not eligible for rich results | Reserve reviews for rated products and software |
| Type without its required properties | Error in the reports, page not eligible | Complete the type's required properties |
| Price or stock out of step with the page | Loss of the enhancement, wrong facts reused elsewhere | Feed the page and the markup from the same CMS field |
Depending on Your Situation
Depending on your starting point, these resources complement this guide:
- your markup varies from one template to another or broke after a redesign: start with a technical SEO audit of your templates;
- you manage a catalogue: connect product markup to your e-commerce SEO strategy;
- your content needs to be better understood by AI systems: align entities and markup with your GEO content strategy;
- you publish studies or webinars: mark them up as Dataset or Event, with the Organization as publisher.
To detect markup errors across all your pages and track your AI presence by keyword, the Incremys GEO/SEO 360 Audit module, a GEO & SEO platform, groups faulty pages by template so you can prioritise fixes with your engineering teams.
FAQ About Structured Data
What Is Structured Data?
It is a normalised representation of information present on a page (product, price, author, address, event), written with the schema.org vocabulary in a strict format. It reduces ambiguity for crawlers, clarifies the page's main entity and can make the page eligible for certain Google rich results, with no guarantee of display.
Does Structured Data Improve Rankings?
No, not directly: Google does not treat it as a ranking factor. It improves understanding of the page and can produce a more informative snippet, which helps indirectly. A thin or poorly targeted page will not climb thanks to its markup; content and search intent come first.
JSON-LD, Microdata or RDFa: Which Format Should You Choose?
Choose JSON-LD in the vast majority of cases. It is the format recommended by Google, separate from the display HTML and simple to generate per template. Microdata and RDFa are still read and supported; they are mainly justified on a site already marked up that way, with stable components, until a migration is planned.
Does Structured Data Belong to Google?
No. Schema.org is an open, community-maintained vocabulary, launched in 2011 by Google, Microsoft and Yahoo, later joined by Yandex, and maintained by a community linked to the W3C. It is read by search engines, social networks, voice assistants and generative engine crawlers. Display in Google is only one of its uses.
Should You Still Mark Up FAQs?
Yes, if the questions and answers are genuinely visible, but not for display. Since August 2023, Google has restricted this rich result to authoritative government and health websites. FAQPage keeps an explicit structuring value for other systems and AI crawlers, with no proven quantified effect on their answers.
Which Pages Should You Start With?
Start with the templates that combine business impact and volume: the homepage with Organization, product or service pages, local pages, and pillar articles with their author. One corrected template covers hundreds of pages at once. Then extend to other page types according to your priorities and the Search Console reports.
How Do You Know Whether Your Markup Is Working?
Check three levels. Syntax and eligibility are verified before going live, on a sample of URLs. Once the pages are indexed, the Search Console Enhancements reports show, by type, which pages are valid, valid with warnings or invalid. Finally, compare the impressions and clicks of the affected pages before and after deployment.
Does Structured Data Work on Multilingual Sites?
Yes, provided each language version has markup that reflects its own visible content: titles, descriptions, prices, availability and contact details in the page's language. Use the URL of each version and keep stable entity identifiers, particularly for the Organization, to avoid contradictions between versions.
.png)
.jpeg)

.jpeg)
%2520-%2520blue.jpeg)
.avif)