Schema Markup for AI Search: What Helps and What Is Hype

The short answer

Schema markup helps search engines and AI systems understand what a page is about, but it does not unlock AI Overviews, AI Mode or ChatGPT citations. Google states that no special structured data is required for AI features. Add accurate markup because it removes ambiguity, not because it forces a result.

That is a more modest claim than you will hear from tools that sell schema as an AI cheat code. It is also the claim that holds up. This guide explains what schema does, which types earn their keep, how to write it, and how to check it.

What schema markup actually does

Schema markup is a vocabulary from schema.org that you add to a page, most often as JSON LD in a script tag. It labels facts in a machine readable way: this is an organisation, this is its phone number, this is an article by this author, this is a product at this price.

Search engines have used it for years to understand pages and to power certain rich results. For AI search, the logic is similar. A system that must work out who a business is, where it operates and what it sells has an easier job when the page states it unambiguously. Microsoft has said publicly that schema helps its language models understand content for Bing and Copilot.

What it does not do is replace the visible content. The text on the page is still what gets read, summarised and quoted. Markup that is not backed by visible text is, at best, ignored.

{
  "@type": "LocalBusiness",
  "name": "Your brand",
  "areaServed": "Dubai",
  "sameAs": [ "linkedin.com/..." ]
}

The check above is a sample layout. It shows the idea of validating that markup matches the visible page. It is not a client result.

What Google says

• There are no additional requirements, and no special schema, to appear in AI Overviews or AI Mode.
• Structured data should match the visible content of the page, and you should not mark up content users cannot see.
• Rich result eligibility depends on Google's current documentation. Some types have been reduced. Google began limiting FAQ rich results in 2023 and, according to industry reports in 2026, has since ended them, and HowTo rich results were removed in 2023.

Those last two points matter. Adding FAQ schema to a service page will not produce a FAQ rich result in Google for most sites. It is still valid vocabulary, and Bing and other tools may read it, but do not sell it internally as a way to take up more space on the results page.

Which schema types are worth adding

Notice what is not on the list: review ratings you cannot show, awards you did not win, or fake addresses. Marking those up is misleading, and can bring a manual action from Google. If your proof is not visible on the page, do not put it in the code.

Step by step: add schema the right way

Step 1. Start from the visible page

List the facts a visitor can see: who you are, what you offer, where you work, who wrote the page, when it was updated. Only these are candidates for markup. If the page does not state something, fix the page first.

Step 2. Choose the type that fits

Pick the most specific type that is true. A dental practice is better described as a Dentist or LocalBusiness than a generic Thing. A service page is a Service, not a Product. When unsure, a plain Organization and WebPage are safe.

Step 3. Write it as JSON LD

JSON LD is Google's recommended format and it is easy to maintain because it lives in one block. Keep names, URLs and contact details identical to those on the page. Here is a minimal organisation example.

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Example Co",
  "url": "https://example.com/",
  "logo": "https://example.com/logo.png",
  "sameAs": [
    "https://www.linkedin.com/company/example"
  ]
}

Step 4. Connect the entities

Link pages together with the same identifiers. Your Organization on the home page should be the publisher referenced in your Articles. Authors can reference a Person. This graph of connected facts helps a system see one consistent business, not many disconnected pages.

The sameAs property lists your official profiles, such as LinkedIn or a verified business listing. Only list profiles you control, and keep the details on them matching your site.

Step 5. Validate

Run the page through Google's Rich Results Test and the Schema Markup Validator. Fix errors, then read the warnings. Remember that passing validation shows the syntax is correct. It does not prove the markup is useful or honest.

Step 6. Keep it in step with the page

The most common failure is drift. A price, phone number or service changes on the page and the markup still shows the old value. Put schema review on your checklist for any page edit. Plugins that generate markup from page fields reduce this risk.

For a visible FAQ, our free FAQ schema generator produces valid JSON LD from your questions and answers. Use it only for questions that really appear on the page.

Common schema mistakes

• Marking up invisible content. If users cannot see it, do not mark it.
• Fake or self written reviews. Never mark up ratings you cannot show and verify.
• Conflicting plugins. Two plugins can output two versions of the same Organization.
• Wrong type. Marking a service page as a Product confuses the page purpose.
• Set and forget. Markup that no longer matches the page loses trust.

How to tell if it is working

Schema rarely produces a single measurable jump, so use indirect checks. Search Console reports on valid structured data and rich result eligibility for the types it supports. Beyond that, track how AI tools describe your business: do they get your name, services and location right? Wrong details are a sign of a messy entity signal that schema and consistent profiles can help repair.

AI visibility overviewSample data

Visibility score58+9 vs last month

Share of voice21%+4 pts

Avg. position2.60.8 better

Net sentiment+41+6

Visibility trendYour brandCompetitor ACompetitor B

By engineVisibility

Google AI Overviews		64	pos 3.1
ChatGPT		55	pos 2.4
Perplexity		60	pos 2.2
Gemini		49	pos 3.0
Copilot		41	pos 3.6

The dashboard above is sample data. Use a fixed prompt set to compare how engines describe you before and after changes, and keep notes of everything else you changed in the same period.

Honest limits

• Schema does not guarantee a rich result, an AI Overview link or a ChatGPT mention.
• Different engines read structured data to different degrees, and they do not publish how.
• Good markup on a thin page is still a thin page.

Want your markup checked against your visible pages? Our technical SEO service includes a structured data review, and the free LLM visibility audit shows how AI tools currently describe you. For the related question of Google features, read how to rank in Google AI Overviews and AI Mode.

Update: FAQ rich results have ended in Google Search

Update on FAQ rich results. Google reduced them in 2023, and its FAQ structured data page now states that the FAQ rich result is no longer shown in Google Search. Its notice says the feature stopped appearing from 7 May 2026, and the page records that the FAQ rich result documentation was removed on 15 June 2026. We read that page in October 2026.

What does that change? Not the case for writing real questions and answers on a page. Visible FAQs still help readers and still give an AI system short, quotable passages. What it changes is the reason. Do not add FAQPage markup to win a result in Google Search. Add it only if the questions and answers are visible on the page, and treat any benefit elsewhere as unproven.

What Google's own guidelines require of any markup

Google's general structured data policies are the rule set that matters most, because breaking them has consequences. Four points from the documentation matter here.

• Markup must describe what users can actually see. Google says not to mark up content that is not visible to readers of the page.
• Markup should be a true representation of the page content.
• Google supports three formats: JSON LD, Microdata and RDFa. JSON LD is the one marked as recommended.
• A structured data issue can lead to a manual action, after which the page loses rich result eligibility. Google says this does not affect how the page ranks in web search, and the Manual Actions report in Search Console is where you check.

For AI features, Google adds one more line. Its page on AI features says there is no special schema.org structured data you need to add to appear in AI Overviews or AI Mode, and that structured data should match the visible text. It also confirms that AI feature traffic is counted in the normal Web search type of the Search Console performance report.

Which schema types does Google document for rich results?

The Search Central gallery lists 25 features at the time we checked it in October 2026, including Article, Breadcrumb, Event, Image metadata, Job posting, Local business, Organization, Product, Profile page, Q&A, Review snippet, Software app, Video and Dataset. FAQ is no longer among them. Two things follow.

1. Many types that are useful for clarity have no Google rich result at all. Service is the clearest example for a service business. It is valid schema.org vocabulary. It does not appear as a Google feature.
2. A type being in the gallery means Google documents a search feature for it, with its own required and recommended properties. A type outside the gallery can still help other systems read your page. We cannot show that it changes Google's behaviour.

Type	Google rich result feature?	Main use for a GEO project
Organization	Yes	State name, logo, contact points and official profiles once, on the home or about page
LocalBusiness and subtypes	Yes	Confirm address, hours, phone and area for a place based business
Article	Yes	Tie guides to an author, a publisher and dates
BreadcrumbList	Yes	Show where a page sits in the site
Product and Offer	Yes	Price, availability and identifiers for a store
ProfilePage	Yes	Mark author or team profile pages that are about a person
Service	No documented feature	Describe the offer, who provides it and where
Person	Used inside other types	Connect writers and experts to their profiles
FAQPage	No longer shown	Valid for visible Q and A, but not a Google lever

Organization markup: what Google actually recommends

Google's Organization documentation says there are no required properties. It recommends adding as many relevant properties as apply. For real world presence, think name, address and telephone. For online presence, think url, logo and sameAs, which lists your profiles elsewhere. It lists others too, such as legalName, foundingDate, contactPoint and identifiers like a tax or VAT number where relevant.

It also gives a practical placement rule: put this on your home page or a single page that describes the organization, such as an about page. You do not need it on every page. The logo image must be at least 112 by 112 pixels and sit at a crawlable, indexable address. Many sites repeat full Organization markup on all pages through a plugin. That is allowed, but it multiplies the work whenever a detail changes, and it is the usual cause of conflicting duplicates.

LocalBusiness markup: required fields and the subtype rule

For LocalBusiness, Google lists name and address as required. It recommends geo coordinates to at least five decimal places, openingHoursSpecification, a telephone number with country and area codes, url and priceRange. It tells you to use the most specific subtype available, and says that if a business fits several types you can list them as an array.

Opening hours trip people up. Google specifies times in hh:mm:ss form, with 24 hour examples like 09:00. Hours that run past midnight go in a single specification. A day closed all day uses 00:00 for both opens and closes. Seasonal changes use validFrom and validThrough in year month day form. If your hours change in Ramadan or on public holidays, update the markup on the same day you update the page.

A connected example: Organization, Service and Article in one graph

The example below is made up. It shows how to link entities using identifiers so that a system sees one business, one service and one author instead of three unrelated blocks. Replace every value with facts that appear on your own pages.

{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Organization",
      "@id": "https://example.com/#org",
      "name": "Example Accounting",
      "url": "https://example.com/",
      "logo": "https://example.com/logo.png",
      "sameAs": ["https://www.linkedin.com/company/example"]
    },
    {
      "@type": "Service",
      "@id": "https://example.com/vat/#service",
      "name": "VAT filing for small businesses",
      "serviceType": "VAT filing",
      "provider": {"@id": "https://example.com/#org"},
      "areaServed": "Dubai"
    },
    {
      "@type": "Person",
      "@id": "https://example.com/team/sara/#person",
      "name": "Sara Example",
      "worksFor": {"@id": "https://example.com/#org"}
    }
  ]
}

The @id values are plain address strings. They give each entity a stable handle, and other blocks refer back with just the @id. Keep an @id the same across pages and across redesigns. If the name or URL of an entity changes, change it everywhere in one pass.

A markup audit you can run in an afternoon

1. Crawl or list your twenty most important pages. Note which have markup and what types.
2. For each page, write down the visible facts: name, service, price range, area, author, update date.
3. Compare each fact with the markup. Mark every mismatch.
4. Look for duplicate entities, for example two Organization blocks from a theme and a plugin.
5. Run each page through Google's Rich Results Test and the Schema Markup Validator. Fix errors first, then read warnings.
6. Check the Search Console enhancement reports for the types you use. Note any that Google has stopped reporting.
7. Record the result with a date. Repeat after every template or plugin update.

Does Bing or Copilot treat schema differently?

Microsoft has said something specific. Search Engine Land reported on 20 March 2025 that Fabrice Canel of Microsoft Bing confirmed, on stage at SMX Munich, that Microsoft uses schema markup to help its language models understand web content. In the same talk he said fresh content matters and recommended IndexNow for pushing new or updated pages. The same article notes it is unclear whether Google's Gemini, OpenAI or Perplexity use schema.

This is a useful, narrow statement. It supports adding accurate markup for Bing and Copilot. It does not tell you how much weight it carries, and it does not generalise to every engine. Our post on Microsoft Copilot, Bing and GEO covers that side in more detail.

Example, hypothetical: a Dubai accounting firm fixes its markup

Imagine a firm whose theme outputs an Organization block with an old phone number, while a plugin adds a second one with the right number and no logo. A page about VAT filing has FAQPage markup for questions that are not on the page. The fix list: remove the duplicate; keep one Organization on the home page with the current number; add Service markup that points to it; delete the FAQPage block or add the visible questions; validate; and write down who updates markup when the phone number changes. No result is promised from any of this. The aim is to remove contradictions a system would otherwise have to guess through.

How schema work fits the GEO Pilot

In the GEO Pilot, structured data is part of the week 2 technical fixes, and the pages we rewrite in week 3 carry schema that matches their visible text. You can draft valid FAQ JSON LD for visible questions with the free FAQ schema generator. For the wider picture on entity clarity, see entity SEO and brand consistency. We test the same prompts again in week 4, and we do not promise that markup changes any answer.