When ChatGPT decides whether to recommend a dentist in Charlotte or an estate attorney in Phoenix, it's not reading your website the way a human does. It's parsing patterns — and structured data is the cleanest, most unambiguous pattern you can hand it. A page without schema forces an AI system to guess: is "Dr. Patel" a person, a clinic, or a brand? Is "5.0" a rating, a price, or a coincidence? A page with proper schema removes the guesswork entirely, which is why structured data has quietly become one of the highest-leverage moves in AI search optimization. It isn't glamorous, and it doesn't drive traffic on its own, but it changes how confidently an LLM can identify, classify, and cite your business when someone asks for a recommendation.
Why schema matters more for AI than for Google
Google has spent two decades getting good at inferring meaning from messy HTML. AI systems, by contrast, are pattern-matching across enormous volumes of text, and structured data is the most reliable signal they get. Where unstructured text can be misread — a bio that mentions a practitioner's alma mater might get attributed to the practice instead of the person — JSON-LD gives an AI system an unambiguous entity type, canonical facts like name, address, phone, and credentials, the relationships between practitioners and the businesses they belong to, and evaluation signals like aggregate ratings and review counts that it doesn't have to infer from prose. For a deeper look at how this fits into the broader optimization picture, our generative engine optimization guide walks through the full stack.
What structured data actually needs to say
The foundation is getting your entity type right, not generic. Schema.org offers vertical-specific subtypes — Dentist, Physician, MedicalClinic, Attorney, LegalService, RealEstateAgent, RealEstateBroker — and the mistake most practices make is defaulting to the generic LocalBusiness type when a more specific one is available. A dentist using the Dentist type instead of LocalBusiness makes it trivially easy for ChatGPT to surface them when someone asks for dentists in their city; the generic type forces inference, the specific type confirms it.
On top of that foundation, FAQPage schema does an unusual amount of work, because AI systems extract question-answer pairs directly from it and frequently quote them back in responses — when someone asks a question your FAQ page already answers, marked up clearly, you've created a citation-ready chunk of content that mirrors exactly how people query AI assistants in the first place. It tends to reward practices that mark up the real questions they already field constantly from clients or patients, not invented ones.
Review and rating data matters for a related reason: when two similar businesses are equally relevant to a query, the one with verifiable aggregate ratings embedded in its schema gives the AI a clean number to anchor confidence on. This doesn't replace your presence on third-party review platforms — it complements it, since AI systems cross-reference your self-declared rating against what they see independently, and consistency between the two builds trust while a contradiction creates doubt.
For named individuals, Person schema links a practitioner to the business they belong to, which is how an AI system resolves a query like "who is the dermatologist in Seattle" down to an actual person rather than a vague practice name. The sameAs property carries particular weight here — pointing to verified profiles on the directories relevant to a given profession tells the model that the person on your site and the person on those other listings are the same entity, collapsing what could otherwise read as several different people into one verified one. Specialty fields go a step further for medical and legal practices specifically: fields like medical specialty, accepted insurance, or declared practice areas let an AI system match a query with several facets — location, specialty, case type — without having to infer any of them from surrounding prose.
Content schema rounds this out. Article or BlogPosting markup with proper author and date fields signals freshness and authority for anything you publish, which AI systems weight more heavily for time-sensitive topics, and it reinforces Person schema at the same time — a managing partner's blog post marked up with both stacks entity signals across multiple pages at once rather than treating them as separate.
A working example
Here's a LocalBusiness block for a dental practice in Charlotte that combines several of these elements:
{
"@context": "https://schema.org",
"@type": "Dentist",
"name": "Queen City Family Dental",
"image": "https://queencityfamilydental.com/images/office.jpg",
"@id": "https://queencityfamilydental.com/#dentist",
"url": "https://queencityfamilydental.com",
"telephone": "+1-704-555-0142",
"priceRange": "$$",
"address": {
"@type": "PostalAddress",
"streetAddress": "1450 Park Road",
"addressLocality": "Charlotte",
"addressRegion": "NC",
"postalCode": "28203",
"addressCountry": "US"
},
"geo": {
"@type": "GeoCoordinates",
"latitude": 35.1956,
"longitude": -80.8487
},
"openingHoursSpecification": [{
"@type": "OpeningHoursSpecification",
"dayOfWeek": ["Monday","Tuesday","Wednesday","Thursday","Friday"],
"opens": "08:00",
"closes": "17:00"
}],
"aggregateRating": {
"@type": "AggregateRating",
"ratingValue": "4.9",
"reviewCount": "287"
},
"sameAs": [
"https://www.healthgrades.com/dentist/queencityfamily",
"https://www.yelp.com/biz/queencityfamilydental"
],
"employee": {
"@type": "Person",
"name": "Dr. Maya Reyes",
"jobTitle": "DDS",
"alumniOf": "University of North Carolina School of Dentistry"
}
}
Every field in that block reduces ambiguity: this is a dentist, in Charlotte, at this address, with this rating, with this dentist on staff, verified across these external sources. That's the whole job schema does.
What schema can't do
Schema is a clarifier, not a creator. It won't manufacture authority you haven't earned — if no third-party source mentions your practice, perfect schema won't get you recommended, because AI systems still need corroborating signals from directories, reviews, and press. It won't override a negative entity signal either: if your real aggregate rating across third-party platforms is well below what you declare in your own schema, the contradiction resolves against you, not in your favor. And it can't substitute for real content — FAQPage schema wrapped around a couple of thin, generic answers won't outperform a genuinely useful, comprehensive resource. Schema amplifies what's already there; it doesn't invent it. If you want to understand how the underlying signals it's amplifying actually get weighted, see our breakdown of how ChatGPT, Claude, Gemini, and Perplexity recommend businesses.
Don't ship broken markup
Before any schema goes live, validate it — the Google Rich Results Test is the industry standard, and pasting in a URL or raw code will flag missing required fields, syntax errors, and warnings, while Schema.org's own validator covers types Google doesn't fully support yet. The errors worth watching for are the ones that quietly undo the whole effort: a missing @context or @type declaration, a name, address, or phone number in schema that doesn't match what's actually on the page, an aggregate rating with no corresponding reviews behind it, or FAQ answers that are too thin or contain markup that breaks parsing. It's worth re-validating after any major site change — a redesign that strips JSON-LD blocks out of your templates is one of the most common, and most avoidable, causes of a sudden AI visibility drop.
How schema fits into your AI visibility score
Structured data is one input, and a powerful one, into the broader question of whether AI systems recognize and recommend your business. The pattern holds up consistently: practices with strong third-party signals but no schema get cited inconsistently, and practices with strong schema but no third-party signals get ignored outright. The real effect shows up when both are in place, and that's when AI systems start naming you by default rather than by accident. To see how schema sits alongside citation density, review signals, and content authority, read our explainer on what an AI visibility score actually measures.
If you're not sure whether your current schema is helping or hurting your AI visibility, the fastest way to find out is to scan your business. RankCommander analyzes how ChatGPT, Claude, Gemini, and Perplexity see your practice — including the structured data signals they're pulling from — and shows you exactly which schema gaps are costing you citations. Run your free AI visibility scan now and get a clear picture of what to fix first.