Structured data for AI search: five schema patterns that actually move citation rates
SE Ranking analyzed citation patterns across thousands of pages appearing in AI search results and found roughly 65% of AI-cited pages include structured data. Wellows put the selection boost from schema markup at around 73%.
These numbers confirm something that SEOs have known for years and AEO practitioners are now rediscovering: when you label your content clearly, the model can read it directly. When you leave the model to infer meaning from unstructured prose, it guesses. Guesses lead to being skipped.
Structured data isn't magic. It's a translation layer between your content and the model's retrieval process. The businesses that get cited consistently are the ones that make the translation easy.
Here are five patterns that move the needle, why each one matters, and what implementation looks like in practice.
Pattern 1: LocalBusiness with the full field set
LocalBusiness is the single highest-leverage schema type for a brick-and-mortar or service-area business. It tells the model exactly who you are and where you operate — the two things that matter most when someone asks "best [service] near me" or "best [service] in [city]."
The minimum viable field set:
{
"@context": "https://schema.org",
"@type": "LocalBusiness",
"name": "Your Business Name",
"address": {
"@type": "PostalAddress",
"streetAddress": "123 Main St",
"addressLocality": "Brooklyn",
"addressRegion": "NY",
"postalCode": "11201"
},
"telephone": "+1-718-555-0100",
"url": "https://yourdomain.com",
"openingHours": ["Mo-Fr 09:00-18:00", "Sa 10:00-16:00"],
"areaServed": ["Brooklyn", "Manhattan", "Queens"],
"sameAs": [
"https://www.google.com/maps/place/your-business",
"https://www.yelp.com/biz/your-business"
]
}
The sameAs field is often skipped and is one of the most valuable. It gives the AI model a set of cross-references — other authoritative sources that corroborate this business's identity. A model can match your structured data against your Yelp listing and your Google Business Profile. That corroboration raises the confidence level that you're a real, well-established business.
Add aggregateRating if you have reviews. A business with 127 reviews averaging 4.8/5 is more citable than one with no review data visible in schema, even if the reviews exist somewhere else on the site. Surface the signal where the model can find it.
For WordPress: the Rank Math or Yoast plugins handle LocalBusiness schema through their Local SEO modules. For Squarespace: inject JSON-LD in the site's code injection header. For Wix: the App Market's SEO tools include schema markup — verify the field coverage before trusting it.
Pattern 2: Service pages with individual Service schema
This is where most small businesses leave the most AI visibility on the table. A single "Services" page that lists everything a business offers gives the model one citation surface. Five individual service pages — each with its own Service schema — give the model five.
The fan-out mechanism in Google AI Mode generates parallel sub-queries for a single user question. "Best home stager in San Francisco" might fan out into sub-queries like "home staging for condos SF," "luxury home staging SF," "cost of home staging SF," and "home staging before-and-after results." A business with one thin service page is invisible to four of those five sub-queries.
The Service schema structure:
{
"@context": "https://schema.org",
"@type": "Service",
"serviceType": "Luxury Home Staging",
"provider": {
"@type": "LocalBusiness",
"name": "Your Business Name"
},
"areaServed": {
"@type": "City",
"name": "San Francisco"
},
"description": "Professional home staging for luxury residential properties in San Francisco and the Bay Area. We specialize in preparing high-value homes for sale.",
"offers": {
"@type": "Offer",
"price": "2500",
"priceCurrency": "USD",
"description": "Starting price for full-room staging consultation and execution"
}
}
The description field should answer the user's question directly and in natural language. This isn't the place for marketing copy. It's the place to state what you do, for whom, and where — the same information the AI model will extract to compose its answer.
Pattern 3: FAQPage schema for question-format queries
AI search queries are often phrased as questions. "How much does it cost to hire a home stager?" "What should I expect from a first staging consultation?" "How long does home staging take?"
A business that directly answers these questions on its site — and formats those answers as FAQPage schema — is speaking the model's native language. The model is already trying to answer a question. You've done the work of converting your knowledge into an answer format it can cite.
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "How much does home staging cost in San Francisco?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Home staging in San Francisco typically costs between $1,500 and $8,000 for a full-service engagement, depending on property size and scope. Most stagers charge separately for consultation, furniture rental, and installation. We offer a free 30-minute consultation to assess scope before providing a custom quote."
}
},
{
"@type": "Question",
"name": "How long does the home staging process take?",
"acceptedAnswer": {
"@type": "Answer",
"text": "For most properties, staging takes 1-2 days for installation after a completed consultation and design phase. Complex or larger properties may require 3-5 days. We typically book staging installations 1-2 weeks after the initial consultation."
}
}
]
}
Note what these answers do: they're specific, include numbers, and preemptively address follow-up questions. An AI model answering "how much does home staging cost" in San Francisco has everything it needs to cite you as a direct, credible source.
Five to eight FAQ entries per service page is a reasonable target. The questions should be drawn from what your customers actually ask — not what you wish they would ask. Look at your email inquiries, your intake form answers, and your Google Business Profile Q&A section for the raw material.
Pattern 4: BreadcrumbList for site structure clarity
BreadcrumbList schema tells AI models how your site is organized — and by extension, how your content relates to your core business identity. It's a low-effort addition that reinforces entity clarity.
{
"@context": "https://schema.org",
"@type": "BreadcrumbList",
"itemListElement": [
{
"@type": "ListItem",
"position": 1,
"name": "Home",
"item": "https://yourdomain.com"
},
{
"@type": "ListItem",
"position": 2,
"name": "Services",
"item": "https://yourdomain.com/services"
},
{
"@type": "ListItem",
"position": 3,
"name": "Luxury Home Staging",
"item": "https://yourdomain.com/services/luxury-home-staging"
}
]
}
This matters most for sites with multiple layers of navigation. A contractor with separate pages for electrical, plumbing, and HVAC services benefits from breadcrumb schema that establishes the relationship between those pages and the parent LocalBusiness entity.
Pattern 5: Review and AggregateRating on the right pages
Reviews are trust signals in AI search, just as they are in traditional search. The difference is that AI models can read structured AggregateRating data directly, without the model having to scrape or infer review sentiment from unstructured text.
Adding AggregateRating to your homepage (or service pages) is straightforward:
{
"@context": "https://schema.org",
"@type": "LocalBusiness",
"name": "Your Business Name",
"aggregateRating": {
"@type": "AggregateRating",
"ratingValue": "4.9",
"reviewCount": "147",
"bestRating": "5"
}
}
A caveat: schema markup ratings need to reflect actual, legitimate reviews. Google and other AI engines cross-reference structured rating data against actual review platforms. Fabricated or inflated aggregate ratings in schema can result in penalties. Use your real numbers from Google Business Profile, Yelp, or industry-specific platforms.
Also worth adding: individual Review schema for selected testimonials that appear on your site. Three or four positive customer reviews formatted as Review schema give the model citable social proof in a machine-readable format.
Implementation priority order
If you're starting from zero:
LocalBusinesswith complete fields andsameAslinks — 60 minutes, highest impactFAQPageon your homepage and top service pages — 90 minutes, addresses question-format queries- Individual
Servicepages withServiceschema — 2-3 hours, expands citation surface area AggregateRatingon homepage — 20 minutes, requires having real review dataBreadcrumbList— 20 minutes, structural clarity for larger sites
Use Google's Rich Results Test (search.google.com/test/rich-results) to validate each schema block before publishing. It shows parse errors, missing required fields, and warnings for non-critical omissions. Fix the errors; warnings are usually fine.
What Autopilot handles
The five patterns above represent a substantial one-time investment in structured data — maybe 6-8 hours for a business starting from scratch. The ongoing work is ensuring those schema blocks stay accurate as your business evolves, and expanding them as AI engines add new supported types.
Lume's Autopilot mode runs the monitoring loop: checking your current schema completeness, flagging fields that are outdated or missing, identifying new service pages that need schema, and surfacing opportunities based on how your category benchmarks against competitors in AI search. The human-hours investment stays small; the citation infrastructure compounds.
See your current schema score: getlumeai.com/search — the free audit shows exactly which structured data fields you're missing and how they affect your AI search visibility.