Structured data for AI: what to mark up first.

Mark up your organization details, your services, and your reviews first. Those three carry the most weight when an AI engine decides whether to name you in an answer. Everything else comes after.

Structured data is code you add to your pages that spells out what things are. A phone number is labeled as a phone number. A review is labeled as a review. A price is labeled as a price. Google has read this markup for years to build rich results. Now AI engines read it too, and they read it because it removes guesswork. A model that has to infer your business hours from a paragraph might get it wrong. A model that reads a schema field for opening hours does not. If you want to be cited by ChatGPT, Google's AI Overviews, or Perplexity, you make your facts machine-readable. That is what structured data does.

What is structured data and why do AI engines care about it?

Structured data is a standard vocabulary that tells machines what each piece of content means. It uses a shared dictionary called Schema.org, and most sites deliver it in a format called JSON-LD, which sits in the page code without changing what a human sees.

AI engines care because they need facts they can trust. When a model answers a question, it pulls from sources it can read with confidence. Plain text forces the model to interpret. Structured data hands it the answer already labeled. A page that says "Open weekdays 9 to 5" in a sentence is a guess. A page with an openingHours field is a fact. Given a choice between guessing and reading a fact, the model reads the fact.

This matters more now than it did five years ago. AI answers compress. A user asks a question and gets one paragraph naming two or three businesses. If your facts are clean and legible, you have a better shot at being one of the names. If a model has to work to understand you, it moves on to a source that made the job easier.

Structured data does not rank your site on its own. It does not fix a slow page or thin content. It makes the good content you already have easier for machines to read and repeat. That is the whole job.

What should you mark up first?

Start with the three schema types that describe who you are, what you do, and whether people trust you. In order: Organization or LocalBusiness, Service or Product, and Review or AggregateRating.

Organization and LocalBusiness schema covers your name, address, phone number, hours, and the areas you serve. For most owner-operated businesses, LocalBusiness is the right choice because it carries location fields. This is the foundation. It tells an engine you are a real business with a real address and a way to be reached. A downtown San Diego dental practice that marks up its name, address, phone, and hours gives every engine a clean record to cite when someone searches "dentist near me open Saturday."

Service or Product schema describes what you sell. A family law firm marks up each practice area as a service: divorce, custody, mediation. A plumber marks up drain cleaning, water heater installation, leak repair. This lets an engine match a specific question to a specific service you offer. Someone asking "who does emergency leak repair in San Diego" is more likely to surface a business that named that exact service in its markup.

Review and AggregateRating schema shows your reputation. AggregateRating carries your average score and review count. Individual Review markup carries specific testimonials. AI engines lean on this heavily because trust signals are hard to fake and easy to read. A firm with a 4.9 average across 200 reviews, marked up cleanly, is a safer name for a model to recommend than one with no visible track record.

Those three cover the questions every buyer asks. Who are you. What do you do. Can I trust you. Answer those in machine-readable form before you touch anything else.

What comes after the first three?

After the core three, add FAQ, Article, and Breadcrumb schema in that order. Each one helps a specific type of content get read and cited.

FAQPage schema marks up your question-and-answer content. This is the closest match to how AI engines work, because they answer questions. If your page already has a clear question and a clear answer, wrapping it in FAQ schema tells the engine exactly where the answer starts and ends. A San Diego immigration firm with a page answering "how long does a green card take" should mark that up. The engine can lift the answer and name the source.

Article schema marks up your blog posts and guides. It carries the author, the publish date, and the headline. This helps engines understand the piece is a considered article rather than a product page or a contact form. Author information matters more as engines weigh who wrote a thing. Marking up your author with real credentials gives your content a signal that thin, anonymous content does not have.

BreadcrumbSchema shows how a page sits inside your site. It maps the path from your home page to the current page. This helps engines understand your structure and helps them present clean navigation in results. It is less urgent than the others, but it is quick to add and it removes ambiguity about how your pages relate.

Do not add every schema type Schema.org offers. Most of them do not apply to a professional services business, and marking up things that are not true or not present on the page causes more harm than doing nothing. Mark up what is real, what is on the page, and what answers a buyer's question.

How do you know if your markup is working?

You test it with the tools the search engines publish, then you watch whether your pages start showing up in richer results. Two free tools do the checking: Google's Rich Results Test and the Schema.org validator.

Paste a page URL or your markup code into either tool. It tells you what schema types it found, whether they are valid, and what errors exist. An error means the engine cannot read that field. A warning means a recommended field is missing. Fix errors first. Warnings are optional polish.

After your markup validates, watch Google Search Console. Under the enhancement reports, Search Console shows which structured data types Google detected and whether any pages have problems. If you added Review markup and Search Console shows it under a review report, Google is reading it. If nothing appears after a few weeks, your markup may be present but not eligible, or the page may not be indexed yet.

The slower signal is citation. AI engines do not report back to you the way Search Console does. You check by asking the questions your buyers ask. Type your service and city into ChatGPT or Perplexity. See if you get named. This is not precise, but it tells you whether the work is landing. Over months, clean markup plus good content plus a fast site adds up to more mentions.

Speed matters here too, because an engine that times out on your page never reads your schema. If your pages load slowly, fix that alongside your markup. We wrote about why a slow site is a sales problem, not an IT problem, and the same logic applies to how AI crawlers treat you. A page that fails to load fast gets read less, cited less, and trusted less. The three numbers behind that are your Core Web Vitals, and they decide whether Google bothers to fully process your page at all.

A short checklist to work through

Run your site against these six items in order and you will have covered the ground that matters.

One. Add LocalBusiness schema to your home page and contact page with name, address, phone, hours, and service area. Two. Add Service schema to each service or practice-area page, naming the exact service. Three. Add AggregateRating and Review schema wherever you display testimonials. Four. Add FAQPage schema to any page with real question-and-answer content. Five. Add Article schema to your blog posts with a named author and a publish date. Six. Validate everything in the Rich Results Test and clear the errors.

Do these in order. The first three do most of the work. If you only ever finish those, you have given every AI engine a clean record of who you are, what you do, and whether people trust you. When we rebuilt the foundation for McShanes Solicitors, clean markup on services and reviews was part of why the right pages started surfacing for the right searches. Structured data was not the whole job. It was the part that made the rest legible.

This work belongs inside your Search Foundations. Markup is one layer of a foundation that also includes a fast site, clear content, and a structure engines can crawl. Add schema to a broken foundation and you have labeled a mess. Add it to a clean one and you have made a good site easy to read. Foundations first.

What structured data does not fix

Structured data does not make thin content worth citing, and it does not rescue a page nobody wants to read. If your service page says nothing useful, marking it up with Service schema only tells the engine your page is empty in a well-organized way. Schema describes what is there. It does not create what is missing. Write the real content first, keep it true, then label it. Markup that describes things not on the page can get your rich results removed, so never mark up a review you do not display or a price you do not honor.