Schema Markup for AI Search: Which Types Actually Matter

Schema markup has been part of SEO for over a decade. Most local businesses either ignore it entirely or implement it incorrectly. For traditional Google Search, the stakes were low. Rich snippets were nice to have, not essential.

For AI search, schema markup is different. It is one of the strongest signals you can control. When AI models crawl your website, structured data gives them machine-readable facts. No interpretation needed. No ambiguity. Just clean data that maps directly to user queries.

But not all schema types matter equally. Some have a measurable impact on AI recommendations. Others are irrelevant. Here is exactly which types to implement, how to do it right, and which mistakes to avoid.

The Schema Types That Matter

Out of the hundreds of Schema.org types, four have a measurable impact on AI recommendations for local businesses. Everything else is either cosmetic (affects Google rich snippets but not AI) or irrelevant.

1. LocalBusiness Schema (Critical)

This is the foundation. LocalBusiness schema tells AI models the essential facts about your business: what you are, where you are, when you are open, how to reach you. Without it, AI models have to guess this information from unstructured page text, and they frequently guess wrong.

Here is a properly implemented LocalBusiness schema:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Plumber",
  "name": "Austin Pro Plumbing",
  "description": "Licensed emergency plumber serving Austin TX. Water heater installation, drain cleaning, leak repair. Same-day service available.",
  "url": "https://austinproplumbing.com",
  "telephone": "+1-512-555-0123",
  "email": "info@austinproplumbing.com",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "4521 Congress Ave",
    "addressLocality": "Austin",
    "addressRegion": "TX",
    "postalCode": "78745",
    "addressCountry": "US"
  },
  "geo": {
    "@type": "GeoCoordinates",
    "latitude": 30.2272,
    "longitude": -97.7631
  },
  "openingHoursSpecification": [
    {
      "@type": "OpeningHoursSpecification",
      "dayOfWeek": ["Monday","Tuesday","Wednesday","Thursday","Friday"],
      "opens": "07:00",
      "closes": "18:00"
    },
    {
      "@type": "OpeningHoursSpecification",
      "dayOfWeek": ["Saturday"],
      "opens": "08:00",
      "closes": "14:00"
    }
  ],
  "priceRange": "$$",
  "areaServed": ["Austin", "Round Rock", "Cedar Park", "Pflugerville"],
  "sameAs": [
    "https://www.google.com/maps/place/?q=place_id:ChIJ...",
    "https://foursquare.com/v/austin-pro-plumbing/...",
    "https://www.yelp.com/biz/austin-pro-plumbing-austin"
  ]
}
</script>

Key details that matter for AI:

• Use the specific @type, not generic LocalBusiness. "Plumber" is better than "LocalBusiness." "ItalianRestaurant" is better than "Restaurant." Specific types create precise category matches.
• The sameAs array is critical. It tells AI models that your Foursquare listing, your GBP, and your Yelp page are all the same entity. This consolidates your signals across platforms.
• areaServed tells AI where to recommend you. Without it, AI may only recommend you for queries mentioning your exact street address city.
• The description field is your pitch to AI. Write it like a direct answer to "what does this business do?" Include services, location, and differentiators.

2. FAQPage Schema (High Impact)

FAQPage schema was the single highest-impact schema type in our testing. Businesses with FAQ schema were not only recommended more often but were recommended with more detailed, accurate information.

The reason is straightforward: FAQ schema provides pre-formatted question-answer pairs. When a user asks AI a question that matches one of your FAQ questions, the AI can pull the answer directly from your schema. No interpretation, no synthesis, just a clean match.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How much does a water heater installation cost in Austin?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Water heater installation in Austin typically costs between $1,200 and $3,500 depending on the type (tank vs tankless) and capacity. Austin Pro Plumbing provides free estimates and includes removal of the old unit in all installation quotes."
      }
    },
    {
      "@type": "Question",
      "name": "Do you offer emergency plumbing service?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Yes. Austin Pro Plumbing offers 24/7 emergency plumbing service throughout Austin, Round Rock, Cedar Park, and Pflugerville. Emergency response time averages 45 minutes. Call 512-555-0123 for immediate assistance."
      }
    }
  ]
}
</script>

The best FAQ questions to include are the ones your customers actually ask. Not marketing fluff. Real questions with real answers. "How much does X cost?" "How long does Y take?" "Do you serve Z area?" "What's the difference between A and B?"

3. Service Schema (Moderate Impact)

Service schema describes specific services your business offers. It creates precise matches between user queries and your offerings. Without it, AI has to infer your services from page text, which is less reliable.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Service",
  "serviceType": "Water Heater Installation",
  "provider": {
    "@type": "Plumber",
    "name": "Austin Pro Plumbing"
  },
  "areaServed": {
    "@type": "City",
    "name": "Austin, TX"
  },
  "description": "Professional water heater installation including tank, tankless, and hybrid systems. Includes old unit removal and disposal. Licensed and insured.",
  "offers": {
    "@type": "Offer",
    "priceRange": "$1200-$3500"
  }
}
</script>

Create a separate Service schema entry for each major service you offer. If you are a plumber, create entries for water heater installation, drain cleaning, leak repair, sewer line replacement, and every other service separately. Each entry is a separate opportunity for AI to match you to a specific query.

4. AggregateRating Schema (Moderate Impact)

AggregateRating makes your review data machine-readable on your own website. While AI models primarily get review data from Google, Foursquare, and Yelp directly, having aggregate rating data on your website reinforces those signals.

"aggregateRating": {
  "@type": "AggregateRating",
  "ratingValue": "4.8",
  "reviewCount": "147",
  "bestRating": "5"
}

This is best added inside your LocalBusiness schema rather than as a standalone block. It tells AI: "This business has 147 reviews averaging 4.8 stars." Simple, clear, machine-readable.

Schema Types That Do Not Matter for AI

To save you time, here are schema types that we tested and found no measurable correlation with AI recommendations:

• Organization - redundant with LocalBusiness for local businesses. Adds nothing AI does not already get from LocalBusiness.
• BreadcrumbList - helps Google understand site structure. AI does not care about your navigation hierarchy.
• WebSite - identifies your site as a website. AI already knows this.
• Article / BlogPosting - useful for content sites. Minimal impact for local business service pages.
• Event - relevant for event venues. Not useful for typical local businesses.
• Product - for e-commerce. Not applicable to service-based local businesses.

This does not mean these schema types are worthless. They help with Google rich snippets and general SEO. But if your goal is AI visibility specifically, focus on the four types that matter and skip the rest until those are perfect.

Common Schema Mistakes That Kill AI Visibility

Implementing schema incorrectly is worse than not implementing it at all. Bad schema confuses AI models and can cause them to present wrong information about your business. Here are the mistakes we see most often.

Mistake 1: NAP Mismatch Between Schema and Platforms

Your schema says your phone number is (512) 555-0123. Your Google Business Profile says 512-555-0123. Your Foursquare listing says 5125550123. These are the same number to humans. They are potentially different entities to AI.

Use the exact same format everywhere. Pick one format and make it identical across your schema, GBP, Foursquare, Bing Places, Yelp, and every other platform. The sameAs property helps, but consistent formatting is the foundation.

Mistake 2: Using Generic @type Instead of Specific

Using "@type": "LocalBusiness" instead of "@type": "Dentist" or "@type": "Plumber" is like telling AI "I am a business" instead of telling it what kind. Schema.org has dozens of specific subtypes. Use the most specific one that applies.

Mistake 3: Missing or Incomplete Address

Leaving out the state, forgetting the zip code, or using abbreviations in one place and full names in another. AI models match businesses to location queries using address data. Incomplete addresses mean missed matches.

Mistake 4: Outdated Hours

Schema says you close at 5 PM but you actually close at 6 PM. The AI tells the user you close at 5. The user shows up at 5:30 and has a bad experience. Worse, the inconsistency between your schema and your GBP hours makes AI less confident about recommending you at all.

Mistake 5: Stuffing Keywords Into the Description

Your schema description should read like a clear, factual summary. Not "Best top-rated affordable emergency plumber Austin TX near me 24/7 plumbing service." AI models are trained to detect keyword stuffing. Write for clarity, not density.

Mistake 6: Duplicate Schema Blocks

Having two different LocalBusiness schema blocks on the same page with slightly different information confuses AI models. One block per type per page. If you have multiple locations, use separate pages with separate schema for each location.

How to Implement Schema Markup

There are three ways to add schema to your website. Each has tradeoffs.

Option 1: JSON-LD in the Head (Recommended)

Add a <script type="application/ld+json"> block in your page's <head> section. This is the cleanest approach. It does not interfere with your visible content, is easy to update, and is the format Google officially recommends.

This is what the code examples above use. Copy them, customize the values, and paste into your HTML. Every page on your site should have LocalBusiness schema. Service pages should add Service schema. FAQ pages should add FAQPage schema.

Option 2: WordPress Plugins

If you use WordPress, plugins like Yoast SEO, Rank Math, or Schema Pro can generate schema automatically. The advantage is ease of use. The disadvantage is that plugins often generate generic schema that misses the specific details AI models need. If you use a plugin, review the generated output and customize it.

Option 3: Google Tag Manager

You can inject schema through GTM, which is useful if you cannot edit your website's HTML directly. This works but adds a dependency. If GTM fails to load, your schema disappears. Direct HTML injection is more reliable.

Testing Your Schema

After implementing schema, test it with these tools in order:

1. Google Rich Results Test (search.google.com/test/rich-results) - validates that your schema is syntactically correct and eligible for rich results. Fix any errors before proceeding.
2. Schema.org Validator (validator.schema.org) - checks compliance with the Schema.org specification. Catches issues the Google tool misses, like deprecated properties.
3. Manual AI Test - ask ChatGPT, Gemini, and Perplexity about your business. Check if the information they return matches your schema. Discrepancies reveal gaps in your implementation.
4. PACO GEO Scan - checks what all four AI models know about your business and identifies schema-related improvement opportunities automatically.

The Schema Implementation Checklist

Here is the minimum schema implementation for a local business that wants AI visibility:

1. Homepage: LocalBusiness schema with complete NAP, hours, description, sameAs links, areaServed, and AggregateRating
2. Each service page: Service schema describing that specific service with pricing, area served, and provider reference
3. FAQ page or section: FAQPage schema with 5-10 real customer questions and detailed answers
4. All pages: Ensure the LocalBusiness name, address, and phone match your directory listings exactly

That is it. Four schema types across your key pages. It takes a few hours to implement correctly. It can take weeks to show up in AI model responses. But once it does, the effect compounds. AI models become more confident in recommending you, which means more recommendations, which means more data supporting those recommendations.

Schema and the AI Data Pipeline

Understanding how schema flows through the AI ecosystem helps explain why it matters:

1. Search engine crawlers (Google, Bing) index your schema as part of their regular crawl cycle
2. Gemini accesses Google's index directly, including your schema data
3. ChatGPT searches Bing, which includes your schema-enhanced search results
4. Perplexity and Claude search the web and encounter your schema in crawled pages
5. AI models with web access can parse your schema in real-time when they visit your site

Schema gives every AI model a clean, consistent data source about your business. Without it, each model is guessing from unstructured text, and each model guesses differently. With it, they all have the same facts, which means more consistent and more frequent recommendations.