Why schema markup has become non-negotiable
For years, structured data was a "nice to have" — a bonus for getting a few stars in Google results. In 2026, the paradigm has shifted. LLMs use structured data to understand the context of your content, not just to display rich snippets.
An analysis by Cludo (Copenhagen, 2025) across 12,000 European sites reveals that sites with complete schema markup are 2.7x more likely to be cited in Perplexity AI than sites without structured data. For Google AI Overviews, the multiplier rises to 3.1x.
The reason is simple: structured data reduces ambiguity. When an LLM crawls your page and finds an Organization block with your name, address, sector and description, it does not need to "interpret" — it knows exactly who you are.
As Jono Alderson, former head of SEO at Yoast (Netherlands), explains: "Structured data is no longer an optional ranking signal. It is the communication layer between your site and any form of artificial intelligence."
The 12 schema types and their priorities
| Schema | Usage | Google impact | AI impact | Priority |
|---|---|---|---|---|
| Organization | Company identity | Knowledge Panel | Very high | P1 |
| FAQPage | Question-answers | FAQ rich snippets | Very high | P1 |
| Article / BlogPosting | Editorial content | Authorship, date | High | P1 |
| BreadcrumbList | Navigation | Breadcrumbs | Medium | P1 |
| WebSite + SearchAction | Site + search | Sitelinks search | Medium | P1 |
| HowTo | Step-by-step guides | Step rich snippets | High | P2 |
| Product | Products / services | Price, reviews, stock | High | P2 |
| LocalBusiness | Local business | Google Maps, local pack | High | P2 |
| Review / AggregateRating | Customer reviews | Stars in SERP | Medium | P2 |
| Event | Events | Event rich snippets | Low | P3 |
| VideoObject | Video content | Video thumbnails | Medium | P3 |
| SpeakableSpecification | Voice search | Google Assistant | High | P3 |
JSON-LD implementation: the method
The format recommended by Google and AI crawlers is JSON-LD (JavaScript Object Notation for Linked Data). It goes in a <script type="application/ld+json"> tag in the <head> of your page, without affecting visible HTML.
The advantages of JSON-LD over microdata and RDFa:
- Separated from HTML — cleaner, easier to maintain
- A single block per schema type — no nested tags everywhere
- Supported by all search engines and AI crawlers
- Easily generated via plugins (Yoast, RankMath) or server-side code
AISOS tip: on a Next.js or Nuxt site, generate JSON-LD server-side (SSR/SSG) to ensure AI crawlers, which do not render JS, can parse it. For FAQ implementation examples, see our article FAQ Structured Data.
The 7 classic mistakes to avoid
According to an audit by Onely (Poland, 2025) across 5,000 European sites, here are the most common errors:
- Orphan schema — a JSON-LD block with no link to the visible content of the page (violation of Google guidelines)
- Outdated information — prices, addresses or hours not updated
- Self-referential schema — a Review on your own page without genuine customer reviews
- Missing @id and @type — entities not correctly identified
- No links between schemas — Organization and WebSite do not reference each other
- Invalid JSON — missing commas, unescaped quotation marks
- Google-only schema — forgetting schemas useful to LLMs (SpeakableSpecification, AboutPage)
Schema markup and AI: the advanced layer
Beyond classic schemas, some implementations specifically target LLMs:
- SpeakableSpecification — indicates the sections of your page most relevant for voice reading or AI citation
- ClaimReview — if you do fact-checking, this schema increases your credibility with LLMs
- DefinedTerm — for glossaries and definitions, extremely cited by AI systems
- Dataset — if you publish original data, this schema makes it referenceable
Andrea Volpini, CEO of WordLift (Rome): "Schemas like SpeakableSpecification and DefinedTerm are still ignored by 95% of sites. It is a huge window of opportunity for those who implement them now."
Find the general technical SEO context in our technical SEO guide 2026 and the integration with AI visibility in our AI visibility guide.
FAQ — Schema markup
Does schema markup directly help Google ranking?
Google states that schema markup is not a direct ranking factor. However, rich snippets improve CTR, and CTR influences ranking indirectly. For AI, schema markup has a direct impact on the probability of being cited.
Which format to choose: JSON-LD, microdata or RDFa?
JSON-LD, without hesitation. It is the format recommended by Google, the easiest to implement and maintain, and the best supported by AI crawlers.
How many schema types can you put on a single page?
As many as relevant. An article page can have BlogPosting + FAQPage + BreadcrumbList + Organization. The important thing is that each schema reflects content actually present on the page.
How do you test your structured data?
Use Google's Rich Results Test and Schema.org's Schema Markup Validator. To check AI compatibility, test with Screaming Frog in "AI crawler" mode (without JS rendering).
Does schema markup work on SPAs (Single Page Applications)?
Yes, provided you use SSR (Server-Side Rendering) or SSG (Static Site Generation). If the JSON-LD is only injected client-side, AI crawlers will not see it.
Do you need to update your schemas regularly?
Yes. The information in your schemas (prices, hours, address, descriptions) must reflect reality. An outdated schema can lead to rich snippet penalties and degrade LLM trust.
Is your structured data up to date?
We audit your schema markup and implement missing types to maximise your rich snippets and AI citations.
Audit my structured data
