Structured data used to be a nice-to-have. Now it's essential.
For years, schema markup was something technical SEO specialists recommended and most businesses ignored. It helped with rich snippets in Google, but it wasn't make-or-break.
That changed when AI started answering questions directly.
Large language models like ChatGPT, Gemini, and Perplexity don't just read your website's content. They read the structure behind it. Schema markup tells AI engines exactly what your business does, what services you offer, where you're located, what your credentials are, and how authoritative you are on specific topics.
Without structured data, AI engines have to guess. And they'll guess wrong, or worse, cite your competitor instead.
What schema markup actually tells AI
Think of schema markup as a machine-readable business card for your website. Here's what you can communicate:
- Organisation schema: Who you are, where you're located, your contact details, social profiles
- Service schema: Every service you offer, with descriptions, pricing, and service areas
- FAQ schema: Direct question-and-answer pairs that AI engines can cite verbatim
- Review schema: Reviews of a product or service that appear on the page itself (not your own business's ratings, which Google does not show for a business reviewing itself)
- Article schema: Author credentials, publication dates, topic authority
- Local Business schema: Operating hours, accepted payments, geographic service area
When an AI engine encounters a website with comprehensive schema markup, it has structured, reliable information to work with. When it encounters a website without it, it has to interpret unstructured text and hope it gets the meaning right.
The direct link between schema and AI citations
Nobody outside the AI companies can see exactly how their engines weigh structured data, but the logic is straightforward: if an AI model needs to answer "what does a bathroom renovation cost in Melbourne?", it's going to prefer a source that has structured pricing data with schema markup over a source that mentions pricing somewhere in a paragraph.
Here's what matters most for AI visibility:
FAQ schema helps answer engines read your answers
FAQ schema presents question-and-answer pairs in a format that AI engines can directly use. When ChatGPT or Google AI Overviews need to answer a question, FAQ schema gives them a pre-formatted, verified answer from your website.
Every business has frequently asked questions. Google now shows FAQ rich results only for a small set of authoritative government and health sites, so the markup will rarely change how your listing looks in search. Its value is that it states each question and answer plainly, in a form any engine can read, as long as the same questions and answers are visible on the page.
Entity authority through schema
Consistent schema markup, with the same name, address, profiles and descriptions everywhere, helps search engines connect your website to your business as one entity. A well-defined entity is easier for AI engines to recognise and to describe with confidence. This is how some businesses get mentioned by ChatGPT when asked for recommendations, while others are completely invisible.
The implementation priority
You don't need to implement every schema type at once. Here's the priority order for AI search visibility:
- Organisation/Local Business: Establish who you are
- FAQ: Answer the questions your customers actually ask
- Service: Define what you do with specificity
- Article: Build topical authority through content
- Review: Only for genuine reviews shown on the page, never your own ratings
What to check before your markup goes live
Most schema problems are not missing markup. They are markup that says something the page does not, or that uses a type the page is not. When we audited our own site in October 2026, most of the structured-data errors came from exactly that: free tools and services described as software applications, a type Google expects to carry user ratings we did not have, and blog posts that described the software that helped write them the same way. None of it was dishonest. The types had simply been chosen without checking what each one requires.
Before any markup ships, check four things.
Does it match the page? Every fact in your structured data should be visible on the page it sits on. If your FAQ schema lists six questions, those six questions and their answers should be readable on the page. If your Service markup includes a price, the same price should appear in the copy. Search engines treat markup that disagrees with the page as a reason to ignore it, and sometimes as a reason to trust the site less.
Is it the right type? Pick the type that describes the page, not the most flattering one. A consulting service is a Service, not a Product. If no type fits, leave it out rather than inventing one: types that schema.org does not define are ignored, and a page full of them looks careless.
Are the reviews real and yours to mark up? Review markup is for reviews of a product or service that appear on that page. Google does not show star ratings for a business marking up reviews of itself, so adding your Google rating to your own homepage schema achieves nothing, and marking up third-party reviews you cannot display risks the opposite.
Does it validate? Paste the code into Google's Rich Results Test and the Schema Markup Validator before publishing, and fix anything either flags. It catches the typos and missing required fields that make a whole block fail silently.
What schema cannot do
Structured data does not make an engine recommend you. It makes your facts easy to read and easy to cross-check. If your website, your Google Business Profile and the directories that list you all say the same thing, schema helps an engine connect them with confidence. If they disagree, schema just states the disagreement more clearly.
It also does not replace the page. An answer engine still needs a clear paragraph it can quote, and a person still needs a reason to call. Treat schema as the label on the jar: worth getting right, but nobody buys the jar for the label.
A practical starting set for a local service business
If you are starting from nothing, this covers most of what matters:
- An Organization or LocalBusiness block on your homepage or contact page, with your name, address, phone number, opening hours and links to your real profiles
- A Service block on each service page, describing that service and the area you cover
- FAQ markup only on pages where the questions and answers are visible
- Article markup on blog posts, with a real author and date
- BreadcrumbList markup if your site has more than one level of pages
Get those right and consistent before adding anything more specialised. One accurate block on every important page does more than a dozen speculative ones on the homepage.
Schema is not optional anymore
Schema markup is the language AI engines speak. The businesses that speak it fluently are the ones getting cited, recommended, and trusted by AI. The ones that don't are hoping AI figures them out on its own.
It won't.
