Schema markup is no longer an SEO optimisation. It is the language you use to speak to AI algorithms.
For years, schema markup was treated as a secondary technical optimisation: something that is “good for SEO”, that might bring stars to the SERPs, useful but not urgent. Developers implemented it if there was time.
In 2026, this approach is costly. Schema markup has become the primary mechanism through which AI systems understand who you are, what you do, who wrote your content and why they should cite you rather than a competitor.
The paradigm shift
| Before (classic SEO) | Now (GEO / AI search) |
|---|---|
| Schema useful for rich results (stars, FAQs, breadcrumbs) | Essential schema for being recognised by LLMs as a trustworthy entity |
| Google reads markup to enrich the SERP visually | AI agents use markup to decide whether to cite a source or ignore it |
| Optional implementation, marginal impact without markup | Without markup, the brand is ambiguous to algorithms, and ambiguous sources do not get cited |
| Format: any (Microdata, RDFa, JSON-LD) | Format: JSON-LD, which Google explicitly recommends for AI-optimised content |
| Validation: nice-to-have | Validation: mandatory, as incorrect markup is ignored or penalised |
How AI agents use structured data
AI agents do not read web pages as a human does. They analyse them as data.
An AI system that has to answer the question “What is the best brand advisory agency in Italy?” does not read every web page from beginning to end. It analyses structured signals to identify the relevant entities, assesses their reliability, and builds the answer by citing the most verifiable sources.
Entity identification, Organization schema tells AI: “This is brand X, it operates in sector Y, it has existed since Z, it is based in W.” Without this, AI has to infer this information from the text, with a greater risk of error and less trust in the source.
Author verification, Author schema says: “This content was written by [nome], who holds these credentials and has this LinkedIn profile.” AI uses this information to assess the source’s expertise — one of the pillars of E-E-A-T.
Answer extraction: FAQPage and HowTo schema structure questions and answers in a machine-readable format. The AI can extract the exact answer without having to interpret the surrounding text.
Reducing hallucinations: schema feeds Google’s Knowledge Graph, which LLMs use as a truth checkpoint. A brand that is well represented in the Knowledge Graph is cited with greater confidence because its data can be verified across multiple sources.
Entity linking: schema with the sameAs property links the brand to its official profiles (LinkedIn, Wikipedia, Wikidata). The AI builds a map of the entity confirmed by multiple sources, and cites it with greater confidence.
The core mechanism: LLMs use schema markup as an anti-hallucination checkpoint. When an AI has to choose between two sources covering the same topic, it prefers the one with verifiable structured data because it lowers the risk of citing incorrect information. Schema is not just metadata; it is a guarantee of accuracy.
The numbers: the impact of markup on AI citability
The data is unequivocal. Markup changes citability measurably.
8.2x: how much more often AI engines cite content with structured data than unstructured content (UpGrowth Schema Research, 2025)
3.2x: more AI citations for sites with correctly implemented schema markup — analysis of 73 sites (Medium / Vicki Larson, 2026)
2.5x: more likely to appear in generative AI answers with appropriate schema markup (Superlines AI Search Data, 2026)
+73%: likelihood of selection in AI Overviews with correct structured schema markup (Wellows, 2026)
+340%: more citations for FAQPage with entity-linked answers compared with equivalent plain text (UpGrowth Schema Research, 2025)
60-90 days: to see measurable improvements in AI citations after full implementation (DemandLocal Agency Analysis, 2026)
The figure that puts everything in perspective: a 30-40% increase in AI visibility with correct schema means that two pieces of content identical in quality and depth, one with complete schema and one without, have AI citation probabilities that differ by 30-40%. This is not a marginal variation. It is the difference between appearing and not appearing in the answers your prospective clients receive every day.
JSON-LD: the correct format
Three formats exist. Only one matters in 2026.
Schema markup can be implemented in three formats: Microdata, RDFa and JSON-LD. The choice is not neutral: since the May 2025 guidelines update, Google has explicitly recommended JSON-LD for content optimised for AI systems.
JSON-LD, RECOMMENDED. Separate from the HTML, it sits inside the script tag. Easy to update without touching the visual markup. Directly readable by AI systems. Google explicitly prefers it for AI optimisation. The only format to use in 2026.
Microdata, NOT RECOMMENDED. Embedded in the visual HTML, every schema attribute must be added to existing HTML tags. Hard to maintain, error-prone, not favoured by AI systems. Not to be used for new implementations.
RDFa, OBSOLETE. Derived from Microdata with a different syntax. It offers no advantages over JSON-LD. More complex to implement. Not optimally supported by modern AI systems. Not to be used.
Basic structure of a JSON-LD
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Bliss",
"url": "https://blissagency.it/en/marketing-communications-brand-advisory-agency/",
"logo": "https://blissagency.it/logo.png",
"description": "Agenzia di brand advisory, governance e operations con sede a Roma e Milano.",
"foundingDate": "2018",
"sameAs": [
"https://www.linkedin.com/company/blissagency",
"https://www.instagram.com/blissagency.it"
]
}
</script>The 10 essential schema types for brands and content
Not all schemas are equal. These 10 are the ones that directly affect AI citability.
01. Organization
For every corporate website, the most important of all.
AI impact: defines the brand as a recognisable entity in the knowledge graph. Without Organization, AI does not know who you are.
Key fields: name, url, logo, description (required); foundingDate, numberOfEmployees, areaServed (strengthen the entity profile); sameAs, an array of URLs of official profiles (LinkedIn, Instagram, Wikipedia, Wikidata); contactPoint with email and telephone (Trust signal); address with addressLocality and addressCountry (local and geographic anchor).
02. Article / BlogPosting
For every article, guide and piece of editorial content.
AI impact: attributes the content to a verifiable author and signals how fresh the content is.
Key fields: headline (the article title); author with @type Person and url to the profile (links the content to the author’s expertise); datePublished and dateModified (recency signal: update dateModified with every revision); image (URL of the main image); publisher (link to the brand’s Organization schema).
03. FAQPage
For Q&A sections and pages with frequently asked questions.
AI impact: the format that generates 340% more citations than the equivalent text. Top priority.
Key fields: mainEntity, an array of Question objects; each Question: name (the question) and acceptedAnswer with text (the answer); the answer must be self-contained and understandable without external context; optimal answer length: 40-80 words; do not use HTML in answers, plain text only.
04. HowTo
For step-by-step operational guides.
AI impact: optimised to answer “how to” queries, both in AEO and in voice search.
Key fields: name (the title of the guide); step, an array of HowToStep with name and text; specificity beats genericity: “7 steps using Figma, 45 min duration” is cited 6.4x more than “7 generic steps”; add tool, duration and cost for each step where applicable; totalTime, total duration of the process in ISO 8601 format (e.g. PT45M).
05. Person (Author)
For every author on the site.
AI impact: links the content to a real person with verifiable credentials, a pillar of E-E-A-T.
Key fields: name, jobTitle, description (full identification); url (link to the LinkedIn profile or the site’s author page); sameAs (verifiable profiles: LinkedIn, Twitter/X, Wikipedia if available); worksFor (link to the Organization schema); knowsAbout, an array of areas of expertise (helps the AI associate the author with the domain).
06. BreadcrumbList
For hierarchical site navigation.
AI impact: helps the AI understand the site architecture and where the content sits within the thematic structure.
Key fields: ListItem in sequence from the homepage to the current page; each item: position (number), name (title), item (URL); implement on every page, not only internal ones; supports understanding of vertical topical authority.
07. WebSite with SearchAction
For sites with internal search functionality.
AI impact: enables the sitelinks search box in the SERP and signals the site structure to the AI.
Key fields: url (site URL); potentialAction, SearchAction with a target that includes the search parameter; query-input: required name=search_term_string. Relevant above all for e-commerce and sites with a large content catalogue.
08. Product / Service
For product or service pages.
AI impact: for those selling physical products or describing services, it does not directly affect citability but it increases rich results.
Key fields: name, description, image (required); offers, PriceSpecification with price and priceCurrency; aggregateRating, ReviewAggregation with ratingValue and reviewCount; areaServed (for services with specific geographic coverage).
09. Dataset
For research, surveys and content with original data.
AI impact: the schema type that unlocks citations for proprietary data, used in vertical search engines.
Key fields: name, description, url (identification of the dataset); creator (link to the Organization or Person that produced the data); datePublished (publication date, critical for recency); variableMeasured (what the data measures). High priority for brands that produce original sector research.
10. SpeakableSpecification
For sections optimised for voice search.
AI impact: signals to Google which sections are suitable for being read aloud.
Key fields: cssSelector, the CSS selector for the sections to flag as speakable; use on direct-answer paragraphs (40-60 words) in the main guides; do not flag technical text, code or content unsuited to being read aloud as speakable; supported by Google Assistant, it increases the likelihood of answering voice queries.
Practical implementation: Organization and Article
The two fundamental schemas, complete examples ready for implementation.
Complete Organization schema
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Bliss",
"legalName": "Bliss S.r.l.",
"url": "https://blissagency.it/en/marketing-communications-brand-advisory-agency/",
"logo": "https://blissagency.it/wp-content/uploads/logo.webp",
"description": "Agenzia italiana di brand advisory, brand governance e brand operations. Specializzata in brand consulting per PMI, gruppi industriali e brand premium.",
"foundingDate": "2018",
"areaServed": ["IT", "EU"],
"address": {
"@type": "PostalAddress",
"addressLocality": "Roma",
"addressCountry": "IT"
},
"contactPoint": {
"@type": "ContactPoint",
"contactType": "customer service",
"email": "hello@blissagency.it"
},
"sameAs": [
"https://www.linkedin.com/company/blissagency",
"https://www.instagram.com/blissagency.it"
]
}
</script>Complete Article schema for a blog post
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "E-E-A-T: la guida completa per capire come Google decide di chi fidarsi",
"image": "https://blissagency.it/wp-content/uploads/eeat-guida-2026.webp",
"datePublished": "2026-04-01T09:00:00+02:00",
"dateModified": "2026-04-18T11:30:00+02:00",
"author": {
"@type": "Person",
"name": "Nome Autore",
"jobTitle": "Brand Strategist",
"url": "https://blissagency.it/team/nome-autore/",
"sameAs": "https://www.linkedin.com/in/nome-autore"
},
"publisher": {
"@type": "Organization",
"name": "Bliss",
"url": "https://blissagency.it/en/marketing-communications-brand-advisory-agency/",
"logo": {
"@type": "ImageObject",
"url": "https://blissagency.it/logo.webp"
}
}
}
</script>FAQPage and HowTo: the two schemas that multiply AEO citability
FAQ and HowTo are the two schemas with the highest impact-to-effort ratio.
FAQPage schema, implementation example
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Cos'è l'E-E-A-T e perché influenza il ranking?",
"acceptedAnswer": {
"@type": "Answer",
"text": "E-E-A-T sta per Experience, Expertise, Authoritativeness e Trustworthiness. È il framework che Google usa per valutare la qualità dei contenuti online. Non è un ranking factor diretto, ma informa i segnali algoritmici che determinano il posizionamento, inclusa la probabilità di essere citati nei risultati AI."
}
},
{
"@type": "Question",
"name": "Quanto tempo ci vuole per migliorare l'E-E-A-T del sito?",
"acceptedAnswer": {
"@type": "Answer",
"text": "In settori non YMYL, i miglioramenti sono visibili in 3-6 mesi con lavoro sistematico su autori, fonti e reputazione esterna. In settori YMYL o molto competitivi il processo richiede 6-18 mesi."
}
}
]
}
</script>HowTo schema, implementation example
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "HowTo",
"name": "Come implementare lo schema markup Organization in JSON-LD",
"description": "Guida step-by-step per implementare il markup Organization sul sito",
"totalTime": "PT30M",
"step": [
{
"@type": "HowToStep",
"name": "Raccogliere i dati del brand",
"text": "Preparare: nome legale, URL, logo, data fondazione, indirizzo, email, profili social ufficiali.",
"url": "https://blissagency.it/guida-schema/#step-1"
},
{
"@type": "HowToStep",
"name": "Creare il JSON-LD nel tag script",
"text": "Aprire il tema WordPress o il template HTML, aggiungere script type application/ld+json nel head.",
"url": "https://blissagency.it/guida-schema/#step-2"
},
{
"@type": "HowToStep",
"name": "Validare con Rich Results Test",
"text": "Incollare la pagina su search.google.com/test/rich-results e correggere eventuali errori prima del deploy.",
"url": "https://blissagency.it/guida-schema/#step-3"
}
]
}
</script>Advanced schemas: Author, Dataset, SpeakableSpecification
The three schema types that make the difference for brands with original content and verifiable expertise.
Author (Person) schema, for author pages
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Person",
"name": "Nome Cognome",
"jobTitle": "Brand Strategist & Head of Advisory",
"description": "12 anni di esperienza in brand strategy per PMI italiane e brand premium. Specializzato in Brand Governance e GEO optimization.",
"url": "https://blissagency.it/team/nome-cognome/",
"image": "https://blissagency.it/team/foto-nome.webp",
"sameAs": [
"https://www.linkedin.com/in/nome-cognome",
"https://twitter.com/nomecognome"
],
"worksFor": {
"@type": "Organization",
"name": "Bliss",
"url": "https://blissagency.it/en/marketing-communications-brand-advisory-agency/"
},
"knowsAbout": ["brand governance", "brand advisory", "GEO", "brand consulting PMI"]
}
</script>Dataset schema, for original research and surveys
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Dataset",
"name": "Survey Brand Governance PMI italiane 2026",
"description": "Analisi condotta su 250 PMI italiane sul livello di governance del brand. Dati su: sistemi di governance attivi, key person risk, presenza schema markup.",
"url": "https://blissagency.it/ricerche/brand-governance-pmi-2026/",
"creator": {
"@type": "Organization",
"name": "Bliss",
"url": "https://blissagency.it/en/marketing-communications-brand-advisory-agency/"
},
"datePublished": "2026-03-15",
"license": "https://creativecommons.org/licenses/by/4.0/",
"variableMeasured": "Livello di brand governance nelle PMI italiane"
}
</script>How to validate schema: tools and processes
Unvalidated markup may be ignored or penalised. Validation is mandatory.
The tools to use are: Google Rich Results Test (search.google.com/test/rich-results), Schema.org Validator (validator.schema.org), Google Search Console (search.google.com/search-console), Structured Data Linter (linter.structured-data.org), and the Schema Markup Validator Chrome extension available on the Chrome Web Store.
The correct validation process
- Develop the JSON-LD locally and validate it with Schema.org Validator.
- Before deployment, test the code directly in Google Rich Results Test.
- Fix all errors, not just warnings, before publishing.
- Post-deployment: check on Google Rich Results Test with a live URL.
- Monitor Google Search Console every 2-4 weeks for new errors.
- Update dateModified in Article schema at every content revision.
The most common mistakes and how to avoid them
Seven errors that make markup invisible or counterproductive.
01. Markup that does not match the visible content. Schema must describe exactly what the user sees on the page. If the markup says “article written by Mario Rossi” but the page shows no author, Google invalidates the markup. Rule: schema specifies, visible content confirms.
02. Missing required properties. Every schema type has required and recommended properties. A FAQPage without mainEntity, or an Article without headline or datePublished, is ignored by rich results. Always check the required properties on schema.org.
03. Incorrect date format. Dates in schema must be in ISO 8601 format: “2026-04-18T09:00:00+02:00”. Dates in Italian (“18 aprile 2026”) or American (“April 18, 2026”) format are not parsed correctly. A common error that invalidates the recency signal.
04. Organization implemented on the homepage only. Organization schema must be present on every page of the site, not just the homepage. AI systems crawl individual pages: if they find an article without Organization in the head, they cannot associate it with the publishing brand.
05. FAQPage with context-dependent answers. The answer in FAQPage must be self-contained and understandable without reading the rest of the page. “As explained in the previous section…” or “See the table above…” are answers that AI systems do not extract. Every answer must stand on its own.
06. sameAs with unverifiable URLs. The profiles linked in sameAs must exist and must mention the brand. A sameAs pointing to an empty LinkedIn profile or a non-existent Wikipedia page weakens rather than strengthens the entity signal. Check every URL before including it.
07. Schema never updated after publication. dateModified in Article schema must be updated every time the content is revised. An article with a 2024 dateModified is treated as outdated by AI systems, even if the content has been updated without touching the markup.
The 6-week implementation plan
How to implement schema markup systematically and measurably.
Weeks 1-2: Foundation
Implement Organization schema across all pages of the site (global header). Implement BreadcrumbList on all internal pages. Validate with the Rich Results Test, with zero errors before proceeding. Audit: which pages already have schema? Which have none?
Weeks 2-3: Content Markup
Implement Article schema on all blog articles (including author and dateModified). Create dedicated author pages with Person schema for each contributor. Update dateModified for every revised article.
Weeks 3-4: FAQ & HowTo
Identify the 10 pages with the most informational traffic. Add an FAQ section at the bottom of each with 5-7 self-contained questions and answers. Implement FAQPage schema on all FAQ sections. Implement HowTo schema on every step-by-step guide.
Weeks 4-5: Advanced & Voice
Implement SpeakableSpecification on the direct-answer paragraphs of the main guides. Implement Dataset schema if the brand has original data to publish. Check and complete the brand’s Wikidata profile where applicable.
Weeks 5-6: Audit & Monitoring
Full audit of all pages with the Chrome Schema Extension. Check Google Search Console for markup errors arising post-implementation. First manual AI citability test: search for the brand on ChatGPT and Perplexity. Set up monthly monitoring with AI citability KPIs.
FAQ
Does schema markup directly affect traditional SEO ranking?
Not directly: Google does not use markup as a direct ranking factor. But it enables rich results (stars, FAQs, HowTo in the SERP), which increase average CTR by 20-30%. And, as documented in this guide, it significantly increases citability in AI results.
Do I need to implement all 10 schema types?
No. The priorities are Organization (on all pages), Article (on all editorial content), FAQPage (on all Q&A sections) and Author (on author pages). The others are added according to the specific content type of the site.
Does WordPress have plugins for schema markup?
Yes, Yoast SEO, RankMath and Schema Pro are the most widely used. They automatically handle Article, BreadcrumbList and some schema variants. But for a complete Organization with sameAs, a custom FAQPage and Author with knowsAbout, you need to add manual JSON-LD to the theme.
How often should I update the schema?
Organization: update if anything changes (address, team, services). Article: update dateModified at every content revision. FAQPage: update if answers change or questions are added. Dataset: update when new data is published.
Does markup work even if the site is not in Google’s top 10?
For Google rich results, markup mainly helps content already on the first pages. For AI citability (ChatGPT, Perplexity), markup is useful regardless of ranking, as these systems also crawl sources outside the top 10. Markup still improves SEO priority over time.