Schema Markup for AI Search: The Structured Data That Gets You Cited

AI Search Visibility Series · Part 7 of 5211 min read

What schema markup actually does (and what it does not)

Every page on your website says something to a human reader. Schema markup says the same thing to a machine, using a shared vocabulary published at schema.org. When your contact page reads “Serving Arlington, Mansfield and Grand Prairie since 2009,” a person understands that instantly. A crawler sees a string of words. Schema turns the string into labeled facts: this is a LocalBusiness, this is its areaServed, this is its foundingDate.

JSON-LD (JavaScript Object Notation for Linked Data) is the format that delivers those facts: a small block of text in the page’s HTML, invisible to readers. Google recommends it, and it is the easiest format to keep consistent across a site.

Here is where honesty matters, because schema gets oversold. Google’s Search Central documentation states there are no additional requirements to appear in AI Overviews or AI Mode and “no special schema.org structured data that you need to add” (Google Search Central, 2025). We quote that to every client. Schema is not a ticket into AI answers.

At the same time, Microsoft’s Fabrice Canel confirmed on stage in 2025 that schema markup helps Microsoft’s large language models understand content (Search Engine Land, 2025), and Copilot runs on Bing’s index. So the accurate framing is this: schema does not earn a citation, but it lowers the odds that a machine misunderstands you, and misunderstanding is the most common reason a qualified business gets skipped or described incorrectly.

One technical point changes how you deploy it. Vercel found that none of the major AI crawlers from OpenAI, Anthropic, Meta, ByteDance or Perplexity render JavaScript (Vercel, 2024). If your JSON-LD is injected by a script after the page loads, Googlebot may see it, but ChatGPT’s crawler will not. Markup has to be in the raw HTML the server sends.

The schema types that matter for a service business

Schema.org lists hundreds of types. A Southlake dental practice or a Fort Worth roofer needs about six, chosen carefully and connected properly.

A note on FAQPage, since clients ask. Google stopped showing FAQ rich results for most sites, so you will not get expandable questions under your listing anymore. That does not make the markup worthless. Google’s documentation on structured data still treats it as a way to help Google understand page content, and answer engines pull from understood content. We keep FAQPage markup on pages with genuine, visible question-and-answer sections, matched word for word.

Article markup deserves more respect than it gets. With AI chatbots now used by about half of U.S. adults (Pew Research Center, 2026), the systems answering those people weigh who wrote something and whether that person has a track record. Article schema with a real author linked to a real Person entity makes that legible. We covered the writing side in Part 6 on answer-first content; schema is how the machine confirms who is speaking.

The connected graph: why stable @id values matter

Most schema on small-business websites is a pile of disconnected fragments. The home page has a plugin-generated Organization block. Each service page mentions the company name as plain text. Blog posts carry an author name with no link to a Person. Nothing references anything else, so the machine has to guess whether “Comfort Air” on the service page is the same “Comfort Air LLC” on the home page.

The fix is a connected graph. Every important entity gets a stable identifier, an @id that looks like a URL, such as https://example.com/#business. That @id never changes, and every other block refers to the entity by it instead of repeating details. The Service block says “provider: this @id.” The Article says “publisher: that same @id.” One truth, referenced everywhere.

Here is a trimmed example for a fictional HVAC company. Notice that the Service block does not repeat the address or phone. It points at the business by @id.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "HVACBusiness",
      "@id": "https://example.com/#business",
      "name": "North Texas Comfort Air",
      "address": { "@type": "PostalAddress", "addressLocality": "Arlington",
        "addressRegion": "TX", "postalCode": "76006" },
      "areaServed": ["Arlington", "Fort Worth", "Mansfield"],
      "sameAs": ["https://www.facebook.com/example"]
    },
    {
      "@type": "Service",
      "@id": "https://example.com/ac-repair/#service",
      "name": "Air Conditioning Repair",
      "provider": { "@id": "https://example.com/#business" }
    }
  ]
}
</script>

The sameAs property is the quiet hero of that block. It tells machines the business described here is the same entity as the one on that Facebook page. That matters because AI systems triangulate. Ahrefs found that branded web mentions correlate with AI Overview brand visibility far more strongly (0.664) than backlinks do (0.218) (Ahrefs, 2025). Mentions only help if the machine can connect them to you, and sameAs draws that line.

The schema mistakes we see most in DFW site audits

Structured Data is where sites most often score low in our assessments for reasons unrelated to effort. The markup is there. It is just wrong, contradictory, or invisible.

• Contradictions between markup, page and profile. Name, address, phone and hours must match across JSON-LD, visible page and Google Business Profile.
• Two plugins both emitting Organization blocks. Duplicate, slightly different entities. Pick one source of truth and switch the other off.
• The generic type when a specific one exists. “Dentist,” “Plumber” or “Restaurant” tells the machine your category without inference.
• Marking up content that is not on the page. FAQ answers that exist only in the schema violate Google’s guidelines and cannot be verified.
• No @id, or @id values that change. If identifiers shift with every plugin update, every reference breaks and the graph collapses into fragments.
• Schema injected by JavaScript. AI crawlers do not run scripts. Markup must be in the server-rendered HTML.
• Self-serving AggregateRating markup. Google prohibits marking up your own business with ratings you collected on your own site. Let the review platforms carry that signal.

How to implement and validate schema in six steps

You do not need a developer on staff, but you do need discipline. This is the sequence our team follows on every engagement.

1. Inventory what is already there. Run key pages through Google’s Rich Results Test and the Schema Markup Validator. Note every block, its source, and any conflicts.
2. Settle your facts first. Confirm the exact name, address, phone, hours, service area, founding year and profile URLs. Fix them on the site and on Google Business Profile before writing markup.
3. Choose types and assign stable @id values. One specific LocalBusiness subtype, one Service per service page, Article for posts, Person for principals, BreadcrumbList on interior pages, FAQPage only where a visible FAQ exists. Document each identifier once.
4. Build the graph. Write the business entity once, then reference it from everything else by @id. Add sameAs links to every official profile.
5. Deploy in the raw HTML. Place the JSON-LD as static markup, not through a tag manager or client-side script, so AI crawlers that skip JavaScript still receive it.
6. Validate, then monitor. Re-run both validators, then check Search Console after every theme or plugin update, since those are the moments schema quietly breaks.

Schema markup helps Microsoft’s large language models understand your content, and pushing fresh content through IndexNow matters because AI systems value recency.

Search Engine Land reporting Bing’s Fabrice Canel, “Microsoft Bing/Copilot use schema for its LLMs”, 2025

One last piece of context. ChatGPT averages about 15 cited sources per response while Gemini averages 3 (Semrush, 2026). Where citation slots are scarce, the sources that win are the ones the model is most confident it understands. Schema is the confidence signal you control completely. For how it fits with content, entities and trust, see our AI search optimization guide.

How AIM helps with schema markup for AI search

Structured Data is one of the four weighted categories in our proprietary AI Visibility Assessment, which scores your site from 0 to 100 alongside Content Authority, Entity Recognition, and Trust & E-E-A-T. The assessment shows what markup exists today, where it contradicts itself, and what a machine cannot confirm about you. From there, our team does the engineering:

• A schema audit that catches duplicate plugins, generic types, JavaScript-only injection and mismatched business facts.
• A connected JSON-LD graph with stable @id values and sameAs links to every official profile, built for your business category.
• Service, Article, Person and FAQPage markup that matches your visible content word for word.
• Monthly monitoring through our AI Visibility Tracker, which reports whether ChatGPT, Perplexity, Gemini and Claude are citing you.

Both tools are included with every AIM SEO plan at no extra charge. Learn more on our AI Search service page.

Schedule a Free Consultation

Frequently asked questions about schema markup for AI search

Find out what the machines think your business is

Our free site test reads your pages the way AI crawlers do. A consultation goes further with a full AI Visibility Assessment, including a structured-data score and a plan to fix what is missing or contradictory.

Schedule a Free Consultation
Call (817) 592-5586

Advanced Integrated Marketing, Inc. · 1905 Ascension Blvd, Suite 154, Arlington, TX 76006 · Serving Dallas-Fort Worth and clients nationwide.

Featured photo by ThisIsEngineering on Pexels.