Schema markup for AI Overviews: what to implement, in what order and how to validate it

Felipe Almeida with Ana Fernández · Updated on September 29, 2026

Ana Fernández ran a test on a large blog she didn't name. Some articles carried author schema and some didn't. When she asked ChatGPT about those authors, all of them well known, it cited the marked-up articles and left the others out. She shared it at SearchBrand's Micro-Training #12, published on May 7, 2026, with no numbers and no written methodology. Read it as field experience from a consultant who has shipped structured data for years, not as a study.

That's the standard for this whole post. Google says no special markup is needed to appear in AI Overviews, and we still recommend schema on every page that drives revenue. Both hold once you split two benefits that most articles blur together: rich results, which is what Google draws on the results page (stars, price, breadcrumbs), and understanding, which is what Google and LLMs can figure out about your page and extract into an answer.

Video chapters

What Google actually says, with dates

We checked Google Search Central documentation on September 24, 2026. Three facts worth keeping handy the next time someone sells you schema as a shortcut:

AI Overviews has no technical requirements of its own. The AI features and your website guide says a page must be indexed and eligible to be shown in Search with a snippet. It asks for no extra markup and no special files. Schema does not guarantee inclusion.
FAQPage rarely produces rich results. On August 8, 2023, Google announced that FAQ rich results would only be shown for well-known, authoritative government and health websites. Ana sees the same in practice: no extra space in Latin America, though she still spots it in some Asian markets.
HowTo rich results are gone. The same announcement limited them to desktop, and in September 2023 Google dropped them there too.

Google's general structured data guidelines add an older rule: markup must describe content users can see on the page. Marking up hidden content can lead to a manual action.

Type	Google rich result (September 2026)	Helps understanding and extraction
Organization	No card of its own; feeds logo and knowledge panel	Yes, defines the brand as an entity
Article / BlogPosting	Better headline, image and date in the result	Yes: author, date, publisher
Product + Offer + Review	Yes: price, availability, stars	Yes
FAQPage	Only well-known government and health sites, since August 2023	Yes, if the questions are visible
HowTo	Retired in 2023	Valid markup, no visual reward
Person / ProfilePage	ProfilePage in forum and perspectives features	Yes: experience and credentials
BreadcrumbList	Yes: breadcrumb trail	Yes: site structure
Service	No rich result exists	Yes
SoftwareApplication	Yes, with price and rating	Yes

No Search Console report confirms the right-hand column. That's why we recommend rolling out in waves and comparing against control pages.

Two layers: the HTML and the JSON-LD

Felipe opened the session with an image (2:40). Google is a bookstore and you drop off a new book every day with no label. The clerk doesn't know which shelf it goes on, and nobody looking for it finds it. Schema is the label.

Ana made it technical at 7:09. Google reads your page twice. First the HTML, with its headings and paragraphs. Then the JSON-LD block, which sits in the <head>, is invisible to visitors and uses the schema.org vocabulary that Google and the LLMs read the same way. She calls it the too long; didn't read of your page: here's the author, here's the date, here's the price.

"Your content can be the best in your industry […] and even so, Google may have no idea what your page is about." — Ana Fernández, founder of Ana Fernández SEO [07:00] (translated from Spanish)

Why would it matter more now? Ana offered a hypothesis and flagged it as her opinion (10:34): AI crawlers behave like Google's crawlers did a few years ago, simpler and "lazier," because they have to fire off many requests fast. Schema hands them the facts pre-chewed. One practical point holds regardless: if your JSON-LD is injected with JavaScript, say through Google Tag Manager, a crawler that doesn't run JavaScript never sees it. Put it in the server-rendered HTML.

She also warned that when you ask Claude or ChatGPT for a schema block, they "invent schema types that then don't work, a million times over." The tool in this post exists because of that sentence.

The author has to be a person

At 14:20 Felipe asked whether the author should be the company or an actual human. Ana's answer was the bluntest of the hour:

"An organization doesn't really write; a person writes." — Ana Fernández [14:40] (translated from Spanish)

Google has long favored people because people can show experience and trustworthiness under E-E-A-T. Ana adds a 2026 argument: with AI anyone can publish 3,000 pieces of content, so volume no longer sets you apart, and a real byline does. In schema, the Article's author is a Person node that points to the Organization via worksFor, and the Organization is the publisher. In-house or freelance doesn't matter. In YMYL topics such as health and finance, the person should have real experience in the field.

Ghostwriters and expert bylines

At 20:39 Felipe raised the case of a bank, where the experts aren't the ones writing the blog. Ana laid out two routes. In the first, the wealth advisor signs what a ghostwriter wrote; she called that "lying" and said it's very common. In the second, the writer signs: someone with 500 finance articles already has authority writing about finance.

We don't recommend the first route. If the expert neither wrote nor reviewed the piece, their byline is exactly the false signal E-E-A-T tries to catch, and in YMYL getting caught is expensive. The honest version: the writer signs, the expert reviews, the page says so ("Reviewed by Dr. X, orthodontist") and the schema declares it with reviewedBy. For health, Ana was stricter: have someone from the field sign it, or pick adjacent topics when no doctor is available.

An author nobody knows

At 47:11 Matías Saavedra asked why you'd mark up an in-house author nobody has heard of. Ana answered with a three-node graph: Organization, Person, and the Article that references both as publisher and author. Authority isn't there on day one. Google gradually ties the name to certain topics, and "by article 15" it shows. An author writing across unrelated industries "isn't ideal."

"Experience, the first E in E-E-A-T, is something only people have." — Ana Fernández [48:10] (translated from Spanish)

Which schema to use for your type of business

At 30:12 Ana ranked types by business model. The last three rows didn't come from the webinar; they're the extension we use at SearchBrand for sectors she didn't cover.

Business	Types, most to least important	Source
B2B and services	Organization, Service, FAQPage, Person	Ana Fernández
Blog or publisher	Article, Person, Organization	Ana Fernández
E-commerce	Product, Offer, Review, BreadcrumbList, Organization	Ana Fernández
SaaS	SoftwareApplication, Organization, FAQPage, Article	Ana Fernández
Healthcare	MedicalBusiness or a subtype (Dentist, MedicalClinic), Person with credentials, Article with `reviewedBy`, FAQPage	SearchBrand
Finance	FinancialService, Person, Article with `reviewedBy`, FAQPage	SearchBrand
Local business	LocalBusiness or a subtype, Service, FAQPage, Person	SearchBrand

For e-commerce Ana stresses Product and Review. One Google detail: since September 2019 it doesn't show stars for reviews a business publishes about itself on its own site. Product reviews remain eligible.

Is FAQPage still worth it?

For rich results, no, unless you run an authoritative government or health site. For understanding, yes: Ana likes it because it packs context and keyword variations into an easy-to-extract format. Use it only on pages that already show a visible FAQ section, with questions people really ask and answers your team has reviewed.

Your brand as an entity: Organization and sameAs

From 31:30 Ana explained why schema moves beyond technical SEO into entity authority. Without markup, Google can't tell that your accounting software and your payroll software belong to the same brand. An LLM may not know who founded the company, which sometimes matters (her example was Patreon), and another company with your name can overshadow you.

Organization fixes that. It lists your founders and contact details, and links your profiles through sameAs. For large companies, Ana said, it can decide whether you get a knowledge panel. Our operating rule: one @id for the organization across the whole site (https://yourdomain.com/#organization), defined on the home page and referenced identically from every page.

How to implement it in four layers

Ana works in four layers (16:40) and made clear it's her method, not the only one.

Map the site. List your page types and how many of each. Everyone skips this; Ana said every time she skips it she ends up doing double the work. Without a map you can't nest schema later.
Prioritize by impact (18:14). Pages that bring traffic and pages that convert go first. Even if a tool writes the code, every change touches the page, so don't do everything at once.
Generate the code by hand, with a library in your stack, or with SearchBrand. If you use an LLM, check every type against schema.org.
Validate and publish (24:26). Run the block through the Schema Markup Validator (validator.schema.org) and Google's Rich Results Test. Then watch the Enhancements reports in Search Console.

If you're starting out, Ana recommends basic schema with one main entity per page connected to the others, and leaving nested schema for when your map is complete.

When to launch the second wave

"Start with 10% of the site and, once you get brave, you can go up to 20%." — Ana Fernández [50:50] (translated from Spanish)

Before wave two, Search Console should show wave one with no errors (22:00). How long to wait depends on how fast Google crawls you. On large, authoritative sites Ana sees the effect almost immediately; on small sites or sites with technical issues she has waited two or three months. She frames it as experiments for a non-technical reason: showing results on a few pages earns stakeholder trust before you ask for more dev hours.

Generate your JSON-LD graph and rollout plan

The generator builds an @graph of four nodes, from Organization down to the page's main entity, linked by @id. It only uses what you type: leave the phone empty and no phone appears. The visibility rule is a hard block: if you mark something as not visible on the page (FAQs, reviews, price, byline), that node isn't generated and the score drops to 0.

The validator checks the JSON-LD you already have live. It compares every @type against a whitelist of 193 schema.org types and suggests the right one when it spots an invented type (BlogArticle → BlogPosting). It flags @id references nobody defines, missing required properties and placeholder values like the classic 555-0123 phone number. Paste the page's visible text and it tells you which schema value doesn't appear there.

The planner takes how many pages you have of each type and returns waves with a publish date, an error check at 7 days and a results review at 4, 8 or 12 weeks depending on crawl speed. From wave 3 on it uses 20% blocks, which is our call, not the webinar's.

JSON-LD schema generator and validator

Your businessBusiness typeSets which types we recommend and in what orderPage typeOne main entity per page

OrganizationBrand nameLegal name (optional)Site URLInclude https://. The #organization @id comes from hereLogo (URL)schema.org subtypeLeave it on "Automatic" if unsurePhoneOnly if it's shown on your site. Leave it empty and we won't make one up

Street addressCityRegion/StatePostal codeCountry

Founders (one per line)Official profiles (sameAs)LinkedIn, Instagram, YouTube, X, Wikidata

Author or personNameJob titleBio page URL on your siteDegrees or credentialsKey for health and financeProfiles (LinkedIn, etc.)Topics they write aboutReviewed by an expert (reviewedBy)Only if the page says so

Reviewer nameReviewer credentials (one per line)

PagePage URLVisible title (H1)Description

PublishedUpdated

Main imageLanguage

Price, exactly as shown on the pageCurrencyAvailability

Average ratingNumber of reviews

FAQs

Paste exactly the ones shown on the page

Paste in bulk

"question | answer" per line, or blocks separated by a blank line.

Breadcrumbs (optional)

What's visible on the page?The FAQs are visible on the pageReviews and rating are visible on the pageThe price is visible on the pageThe author is named on the page

Score

100/100

Ready to validate

Warning (1)

Dentist without an address. We never make one up: if you serve customers on site, enter the real address shown on your website.

{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Dentist",
      "@id": "https://clinica-dental.example/#organization",
      "name": "Clínica Dental Ejemplo",
      "url": "https://clinica-dental.example/",
      "logo": {
        "@type": "ImageObject",
        "url": "https://clinica-dental.example/logo.png"
      },
      "sameAs": [
        "https://www.instagram.com/clinicadentalejemplo",
        "https://www.linkedin.com/company/clinicadentalejemplo"
      ]
    },
    {
      "@type": "WebSite",
      "@id": "https://clinica-dental.example/#website",
      "url": "https://clinica-dental.example/",
      "name": "Clínica Dental Ejemplo",
      "inLanguage": "es-CL",
      "publisher": {
        "@id": "https://clinica-dental.example/#organization"
      }
    },
    {
      "@type": "Person",
      "@id": "https://clinica-dental.example/#author-dra-camila-rojas",
      "name": "Dra. Camila Rojas",
      "jobTitle": "Ortodoncista",
      "url": "https://clinica-dental.example/equipo/camila-rojas",
      "sameAs": [
        "https://www.linkedin.com/in/camila-rojas-ejemplo"
      ],
      "knowsAbout": [
        "ortodoncia invisible",
        "alineadores transparentes"
      ],
      "hasCredential": [
        {
          "@type": "EducationalOccupationalCredential",
          "name": "Cirujana dentista"
        },
        {
          "@type": "EducationalOccupationalCredential",
          "name": "Especialista en ortodoncia"
        }
      ],
      "worksFor": {
        "@id": "https://clinica-dental.example/#organization"
      }
    },
    {
      "@type": "FAQPage",
      "@id": "https://clinica-dental.example/ortodoncia-invisible#webpage",
      "url": "https://clinica-dental.example/ortodoncia-invisible",
      "name": "Ortodoncia invisible con alineadores transparentes",
      "isPartOf": {
        "@id": "https://clinica-dental.example/#website"
      },
      "inLanguage": "es-CL",
      "description": "Tratamiento de ortodoncia con alineadores transparentes: precio, duración y financiamiento.",
      "dateModified": "2026-09-24",
      "reviewedBy": {
        "@id": "https://clinica-dental.example/#author-dra-camila-rojas"
      },
      "mainEntity": [
        {
          "@type": "Question",
          "name": "¿Cuánto cuesta la ortodoncia invisible en Chile?",
          "acceptedAnswer": {
            "@type": "Answer",
            "text": "Depende de la complejidad del caso y del número de alineadores que necesites. Lo definimos después de una evaluación con escáner intraoral y te entregamos el presupuesto por escrito en la primera consulta."
          }
        },
        {
          "@type": "Question",
          "name": "¿Qué es mejor, alineadores o brackets?",
          "acceptedAnswer": {
            "@type": "Answer",
            "text": "Los alineadores se sacan para comer y casi no se notan, pero exigen usarlos unas 22 horas al día. Los brackets no dependen de tu disciplina. En la evaluación te decimos cuál conviene para tu mordida."
          }
        },
        {
          "@type": "Question",
          "name": "¿Cuánto dura el tratamiento?",
          "acceptedAnswer": {
            "@type": "Answer",
            "text": "Un caso leve puede terminar en seis meses y uno complejo tardar más de un año. El plan digital que te mostramos en la evaluación incluye la duración estimada para tu caso particular."
          }
        },
        {
          "@type": "Question",
          "name": "¿Qué tipos de alineadores existen?",
          "acceptedAnswer": {
            "@type": "Answer",
            "text": "Hay alineadores de distintas marcas y planes según la cantidad de etapas. La diferencia principal está en cuántos alineadores incluye el plan y en los controles, no en el material."
          }
        },
        {
          "@type": "Question",
          "name": "¿Se puede pagar en cuotas?",
          "acceptedAnswer": {
            "@type": "Answer",
            "text": "Sí. Puedes pagar con tarjeta de crédito en cuotas o con un plan de pagos mensuales de la clínica. Las condiciones quedan escritas en el presupuesto antes de empezar el tratamiento."
          }
        }
      ]
    }
  ]
}

Recommended types for your business

1. MedicalBusiness SearchBrand
Google rich result: No result card of its own; Google may use logo and details for the knowledge panel. Key for entity understanding.
Understanding and extraction: yes
2. Person SearchBrand
Google rich result: No Google rich result. Useful for understanding and extraction.
Understanding and extraction: yes
3. Article SearchBrand
Google rich result: Enhances article appearance (headline, image, date, author). No guaranteed visual feature of its own.
Understanding and extraction: yes
4. FAQPage SearchBrand
Google rich result: Rich result limited since August 2023 to well-known government and health sites. Still useful for understanding if the questions are visible on the page.
Understanding and extraction: yes

Before you publish

Paste the block in this URL's <head> (in site builders: header "custom code").
Validate at validator.schema.org and in Google's Rich Results Test.
Check Search Console Enhancements reports after 7 days.
Update the schema when address, phone, prices or profiles change.

Check a live URL

Paste your JSON-LD (with or without <script>)Visible page text (optional)Copy it from the browser with Ctrl+A. We use it to spot schema values that don't appear on the page

The FAQs are visible on the page

Reviews and rating are visible on the page

The price is visible on the page

The author is named on the page

Whitelist: 193 types (schema.org v29 · SearchBrand 2026-09-24).

Score

0/100

6 errors

Error (6)

"Company" does not exist in schema.org. Use "Organization".@graph[0]
"BlogArticle" does not exist in schema.org. Use "BlogPosting".@graph[1]
@id "https://clinica-dental.example/#autor" is referenced but no node in this block defines it. If it lives on another page (e.g. the home page), use exactly the same @id URL there.@graph[1].author
"telephone": "+56 9 1234 5678" looks like a placeholder or made-up value. Use the real value or remove the property.@graph[0].telephone
"streetAddress": "Calle Falsa 123" looks like a placeholder or made-up value. Use the real value or remove the property.@graph[0].address.streetAddress
"datePublished": "24/09/2026" is not ISO 8601 (YYYY-MM-DD).@graph[1].datePublished

Warning (11)

@id "#org" is not an absolute URL. Use https://yourdomain.com/#name so it stays stable across pages.@graph[0]
PostalAddress missing addressCountry (recommended).@graph[0].address
FAQPage with a single question.@graph[2]
Product missing image (recommended).@graph[3]
FAQPage present, but you marked that the page shows no FAQs.
Review or AggregateRating present, but reviews are not visible on the page.
telephone "+56 9 1234 5678" does not appear in the visible text you pasted.@graph[0].telephone
streetAddress "Calle Falsa 123" does not appear in the visible text you pasted.@graph[0].address.streetAddress
Question "¿Duele?" does not appear in the visible text you pasted.@graph[2].mainEntity[0].name
ratingValue "4.9" does not appear in the visible text you pasted.@graph[3].aggregateRating.ratingValue
ratingCount "212" does not appear in the visible text you pasted.@graph[3].aggregateRating.ratingCount

Info (1)

Rich result limited since August 2023 to well-known government and health sites. Still useful for understanding if the questions are visible on the page.

Detected types

CompanyPostalAddressBlogArticleFAQPageQuestionAnswerProductAggregateRating

Defined @ids	Referenced @ids	Orphan @ids
#org	https://clinica-dental.example/#autor	https://clinica-dental.example/#autor

Page typeNumber of pagesPriority (high = drives traffic or converts)Is home page

Is home page

Start date

Crawl speedFast (large site, 4 weeks)Normal (8 weeks)Slow (small or new site, 12 weeks)

Advanced

First wave %Cumulative second wave %% per wave from wave 3

29 pages in total · results read at 8 weeks

Wave	Pages by type	Total	Cumulative %	Publish	Check errors	Read results
1 Webinar	Home 1 · Tratamientos 2	3	10.3%	2026-09-28	2026-10-05	2026-11-23
2 Webinar	Tratamientos 3	3	20.7%	2026-11-23	2026-11-30	2027-01-18
3 SearchBrand	Tratamientos 1 · Blog 5	6	41.4%	2027-01-18	2027-01-25	2027-03-15
4 SearchBrand	Blog 6	6	62.1%	2027-03-15	2027-03-22	2027-05-10
5 SearchBrand	Blog 6	6	82.8%	2027-05-10	2027-05-17	2027-07-05
6 SearchBrand	Blog 3 · Profesionales 2	5	100%	2027-07-05	2027-07-12	2027-08-30

Timeline

2026-09-281
2026-11-232
2027-01-183
2027-03-154
2027-05-105
2027-07-056

Info (1)

With 29 pages, 10% is 3 URLs: any traffic change will be a handful of clicks and hard to tell from noise. Compare against control pages.

FAQs are only as good as the questions you pick. To see what people ask about your category in ChatGPT or AI Overviews before you write them, SearchBrand gives you 10 free credits, no credit card.

Worked example: a dental clinic from zero schema to a connected graph

This example is illustrative. It follows Felipe's demo from 38:00, built on an orthodontics clinic whose site ran on a GoHighLevel-based page builder. The page counts are ours.

Map. The clinic has 29 URLs: home, 6 treatment pages, 20 blog posts and 2 practitioner profiles.
Prioritize. 10% means 3 URLs: the home page, where the Organization node lives, plus the 2 treatments with the most Search Console impressions.
Research the questions. In the demo, Felipe used SearchBrand to see what people ask about clear aligners: cost, aligners versus braces, which types exist. Add treatment length and financing.
Publish the FAQs on the page, with answers reviewed by the orthodontist. It's YMYL: no promised outcomes.
Build the graph. Dentist with @id #organization, logo and sameAs; phone and address only if they're the real ones shown on the site. Person for the orthodontist with credentials and worksFor. The page as FAQPage with the five visible questions and reviewedBy pointing to her. This is the tool's default state.
Validate at validator.schema.org and in the Rich Results Test. It will tell you FAQPage won't earn a rich result for a private clinic; that's expected.
Insert. Felipe pasted the block into the page header's custom code and hit publish. Wix or Shopify ship basic schema; this adds to it.
Measure. Publishing on September 28, 2026 means checking Search Console Enhancements on October 5. On November 23, compare impressions and CTR for those 3 URLs against 3 control pages and check whether the clinic shows up in AI answers about the treatment.
Decide on wave 2: no errors and some signal means moving to 20%, which is 3 more treatment pages.

It helps to know what each URL has today. At 43:40 SearchBrand showed a free checker (schematic-sb.lovable.app) that reads a single page, not the whole site. On anafernandezseo.com it found only contact schema. Ana laughed it off as the shoemaker's children going barefoot.

What we don't recommend

Asking an LLM for schema and pasting it unchecked. Most of the errors Ana listed from 25:37 start there: blocks copied from another site with phone numbers that never belonged to the business, or a Product type on a blog post.

Declaring what the page doesn't show. Ana compared it to buying a book labeled magical realism "and you open it and it's chicken recipes."

"If the rating isn't on your page, please don't put it in the schema. It might work at first, and it can get you penalized later." — Ana Fernández [35:20] (translated from Spanish)

She noted that in low-competition markets it sometimes slides. In regulated industries it doesn't.

Nesting schema in month one. Blocks that contradict each other across pages cause more errors than a simple, well-connected graph.

Forgetting it after launch. Ana finds sites with a three-year-old address and closed X profiles in sameAs. Schedule a review whenever contact details change.

Does schema speed up indexing?

No. Ana was clear at 52:39: a new site has to wait and earn mentions and links, without buying them in bulk. If the site isn't new and Google is still slow, look at architecture and JavaScript dependency, since Google renders JavaScript in a second pass. IndexNow helps with Bing. "Slow indexing is a symptom of something bigger."

How to tell if it worked

Ana called this "the absolute biggest pain" in SEO (55:07). In e-commerce you compare product-page CTR before and after. On a blog you'll see more traffic, and the next question is whether it brings revenue; that takes attribution work in analytics.

Her strongest case: a finance site with roughly 1 million monthly visits that, by her account, grew traffic 15% after properly implementing Article and Person schema tied to the Organization. She didn't give the period or the name. She also supplied the counterweight: on a small site that same 15% might be "two clicks," and then it's hard to separate the effect from luck.

For small sites we suggest comparing the wave's URLs against a control group with similar impressions over the same 8-week window. On the AI side, track how often your brand gets cited for the questions in your category; our AEO and GEO audit guide covers how.

What about llms.txt?

Fernando asked at the very end (57:37). Ana said she analyzed traffic with and without llms.txt for an article she wrote for Search Engine Land and found no difference. Google's AI features guide doesn't ask for new machine-readable files either. Our position: there's no public evidence it moves traffic, it takes an afternoon to set up, and it shouldn't compete with schema for priority. Keep it if you have it; otherwise start with Organization and Person.

Where to start this week

Export the Pages report from Search Console for the last 3 months and count your URLs by type. Put those numbers into the planner. If 10% comes out to 3 URLs, make them your home page and the 2 that convert best, and leave the blog for later. Check errors at 7 days and don't touch wave 2 before the review date.

To find out whether those pages start showing up when people ask ChatGPT or AI Overviews about your category, SearchBrand tracks your brand's mentions and citations. You can start with 10 free credits, no credit card.

Sources

Webinar "Schema y datos estructurados: cómo posicionarte en AI Overviews," SearchBrand Micro-Training #12 with Ana Fernández (in Spanish). YouTube, May 7, 2026.
Google Search Central, AI features and your website: developers.google.com/search/docs/appearance/ai-features
Google Search Central Blog, Changes to HowTo and FAQ rich results, August 8, 2023.
Google Search Central, General structured data guidelines.
Google Search Central Blog, Making Review Rich Results more helpful, September 16, 2019.
Schema.org full type hierarchy: schema.org/docs/full.html

Frequently asked questions

Does schema help you appear in AI Overviews?

It helps Google understand your page, but it doesn't guarantee inclusion. Google Search Central says a page needs to be indexed and eligible to show with a snippet to appear in AI Overviews and AI Mode, with no special markup. Schema makes the author, date or price easier to extract, as long as it matches what the page shows.

Which schema should I implement first?

Organization on the home page, with a single @id the whole site references. Then, by business model: Article and Person for blogs and publishers; Product, Offer and Review for e-commerce; Service and FAQPage for B2B services; SoftwareApplication for SaaS. Ana Fernández recommends one main entity per page, connected to the others, before moving to nested schema.

Should an article's author be a person or the company?

A person. Google looks for experience and trustworthiness, and people demonstrate those. In schema, the author is a Person node with job title, bio and sameAs, linked to the Organization via worksFor, while the company is the publisher. In health or finance, if the writer isn't an expert, add a credentialed reviewer shown on the page and declared with reviewedBy.

Is FAQ schema still worth it in 2026?

For rich results, barely: since August 8, 2023, Google only shows them for well-known government and health sites. For understanding, yes, because it delivers questions and answers that are easy to extract. Use it only when the page displays those questions and they answer real doubts. HowTo is worse off: Google retired its rich results in 2023.

How do I validate my structured data?

Run the block through the Schema Markup Validator (validator.schema.org), which checks the vocabulary, and Google's Rich Results Test, which tells you whether it's eligible for rich results. Seven days after publishing, check the Enhancements reports in Search Console. Also confirm by hand that phone, address, price and reviews match what's visible on the page.

How many pages should I mark up first?

Ana Fernández recommends starting with 10% of the site and moving to 20% once you've confirmed there are no errors. Pick the home page and the pages that bring the most traffic or conversions first. If Google crawls your site quickly, you can read results in about 4 weeks; on small or slow-crawled sites, give it 2 or 3 months.