Schema Markup for AI Search in 2026: The Types That Matter and the Ones That Do Nothing
By Alexey Chasin · Updated September 25, 2026 · 9 min read
Schema markup is structured data on a page, written in JSON-LD, that tells search engines and AI systems exactly what the page is about and who is behind it. In 2026 seven types earn their place on a local business site: LocalBusiness or a subtype with sameAs links to the Business Profile and directories, Service for each service page, FAQPage for the questions, Article or BlogPosting with a named author for guides, Offer with a price range where prices exist, BreadcrumbList, and Speakable on the answer paragraph. Schema does not make a weak page rank or get cited. It makes a strong page unambiguous, which is what an AI deciding whether to trust and attribute a source needs.
What schema does in AI search
AI engines reason about entities: this business, in this place, offering these services at these prices, written by this person. A page's words say that loosely. Its schema says it exactly, in a format that never needs interpreting. Every 2026 guide from Frase to Writer to The HOTH lists schema as a shared requirement across SEO, GEO and AEO, and Google's own guidance says AI Overviews and AI Mode use the same structured data as Search. Schema is not a ranking trick. It is how a page becomes a clean fact instead of a paragraph to interpret.
The seven types that matter
| Type | Where | What it must contain | Why AI uses it |
|---|---|---|---|
| LocalBusiness or subtype (ProfessionalService, RoofingContractor, Dentist) | Every page, once | Name, phone, address or areaServed, hours, priceRange, sameAs links to Business Profile, LinkedIn, Yelp, directories | Identifies the entity and ties it to the profiles AI already trusts |
| Service | Each service page | Name, description, provider, areaServed, offers | Lets an engine match a question to the exact service |
| FAQPage | Pages with an FAQ | Question and Answer pairs matching the visible text | Direct question-answer pairs are the easiest thing to extract and cite |
| Article or BlogPosting | Guides | Headline, author as a Person, publisher, datePublished, dateModified | Attributes the content to a person and a date; freshness and authorship both weigh |
| Offer with priceSpecification | Pages with prices | Price or price range, currency | Prices are what AI answers and agents extract first |
| BreadcrumbList | Every page | The page's position in the site | Shows an engine the site's structure and the page's topic |
| Speakable | Answer paragraph and FAQ | CSS selectors of the sentences to read aloud | Marks the passage assistants should quote |
The types that do nothing
- Review and AggregateRating on your own site for your own business. Google ignores self-serving ratings for LocalBusiness rich results, and AI engines read the Business Profile and Yelp instead.
- HowTo for pages that are not step-by-step instructions. Google removed HowTo rich results in 2023; the markup still describes content but earns nothing for a service page.
- Organization on every page when LocalBusiness is already there. Pick one entity type for the business and use it consistently.
- Keyword-stuffed 'about' and 'knowsAbout' fields. Engines read them as text and discount padding.
- Any schema that describes content not visible on the page. Google's guidelines treat that as spam, and it is the fastest way to lose the rich results you do have.
sameAs is the line that ties it together
The single most useful line in a LocalBusiness block is sameAs: the URLs of the Business Profile, LinkedIn, Yelp, Clutch or the directory for your trade, and the owner's LinkedIn on the Person block. It tells every engine that the business on this site, the one on Google Maps and the one with 40 Yelp reviews are the same entity. Without it, three entities with the same name compete for trust. With it, the reviews and listings AI already trusts vouch for the site.
How to check yours
- Google's Rich Results Test shows what Google parses and any errors.
- The Schema Markup Validator at validator.schema.org checks the JSON-LD itself.
- View the page source and confirm the details match the visible page and the Business Profile exactly: same name, same phone, same hours.
- Search Console's Enhancements reports list pages with FAQ, breadcrumb and other markup and their errors.
What schema cannot do
It cannot make a page that does not answer the question get cited. It cannot substitute for a Business Profile or reviews. It cannot fix a page that takes five seconds to load or is blocked to crawlers. And it cannot be added once and forgotten: when the phone number, hours or prices change, the schema changes with them or it becomes a lie the engine catches. On a hand-coded site the schema is generated from the same data as the visible page, so it cannot drift. On a plugin-driven site it usually does.
Every page on this site carries the seven types above, generated from one data source so the visible text and the markup never disagree. View the source of any page to see the pattern.