SEO / AEO

Using Structured Data to Help AI Answer Engines Understand a Business

What JSON-LD actually does in 2026, and why the same schema.org shapes that helped Google now help the answer engines too.

Published 2026-03-04 · By Claire Miller

A typical small business website in 2026 has dozens of factual claims about the business scattered across pages, some accurate, some stale, some duplicated, some implied. An answer engine trying to cite the business cannot rely on inference from prose alone. It needs the facts in a form it can read confidently. That form is structured data, and the schema is schema.org.

What structured data does

A page's prose tells a person what the business does. A page's JSON-LD structured data tells a machine the same thing, with the entity types and properties it can read directly. For a search engine or an answer engine, the structured data is the canonical map of the page: name, address, services, prices, hours, products, reviews, FAQ pairs, organizational parents. The prose is decoration around it.

What changes in 2026 is the audience. In 2018, the audience was Google. In 2023, the audience was still mostly Google plus Bing. In 2026, the audience is every answer engine that reads the page: Perplexity, ChatGPT search, Google's AI Overviews, Bing Copilot, Apple's web-crawler-assisted Siri, and the long tail of specialized assistants. The same JSON-LD you wrote for Google in 2019 works for them now because they all read schema.org the same way.

What schema types matter for small businesses

Not all schema types matter equally. For most small businesses, six types carry 95 percent of the value:

Organization. The business itself. Name, logo, URL, contactPoint, sameAs (links to the canonical social profiles), foundingDate if known.

LocalBusiness. A subtype of Organization for businesses that serve a geographic area. Address, geo (latitude/longitude), openingHoursSpecification, areaServed, priceRange, paymentAccepted, currenciesAccepted. Use the most specific LocalBusiness subtype applicable: Dentist, Attorney, Plumber, Electrician, RealEstateAgent, Restaurant, Store, etc.

Service. A service the business offers. serviceType, provider (refs the LocalBusiness), areaServed, offers (a price specification, hours, etc.), description.

FAQPage. Frequently Asked Questions, with mainEntity (a list of Question items, each with acceptedAnswer). Visible to engines and useful in AI citations.

Product. A specific product or SKU. name, image, description, brand, offers, aggregateRating.

Review or AggregateRating. Standalone reviews or aggregated ratings from a review system. Required to wire up AggregateRating on a LocalBusiness or Product.

The long tail of schema types is real but optional for a small business starting from zero. Get these six right before adding BreadcrumbList, WebSite, VideoObject, or the more specialized types.

Where to put the JSON-LD

Every page that asserts something factual about the business has JSON-LD in its head or at its top of body. The simplest implementation is one script block per page; the cleaner implementation uses a generator plugin (Astro schema-dts, Eleventy eleventy-plugin-schema, Hugo templates) that emits the right script from the page's data file.

For a small business, the right answer is the generator plugin. Static-site generators make it cheap to centralize the JSON-LD generation in a layout file. The discipline is that every page that should have JSON-LD does, and every page that should not have JSON-LD does not.

What not to bother with

Three categories of structured data are not worth a small business's time in 2026:

Custom JSON-LD for "AI crawler" protocols. The various "llms.txt"-style proposals and similar concepts do not have meaningful market weight. Schema.org's well-supported types are the right focus.

Keyword stuffing inside structured data. Adding the same keywords three times across service, FAQ, and reviews schema does not help; it hurts. Be accurate, be brief.

Schema that is not visible on the page. Google's structured data guidelines require that the entity's content be visible on the page; the schema describes what the page already says. Schema that invents facts is misleading and is the canonical way to get manual actions against a site.

A working example

For a fictional plumbing company in 2026, the Organization + LocalBusiness + Service JSON-LD block on the contact page reads roughly:

{
  "@context": "https://schema.org",
  "@type": "Plumber",
  "name": "Acme Plumbing",
  "url": "https://acmeplumbing.com/",
  "telephone": "+1-555-0142",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "123 Main St",
    "addressLocality": "Springfield",
    "addressRegion": "IL",
    "postalCode": "62701",
    "addressCountry": "US"
  },
  "geo": { "@type": "GeoCoordinates", "latitude": 39.7817, "longitude": -89.6501 },
  "openingHoursSpecification": [
    { "@type": "OpeningHoursSpecification", "dayOfWeek": ["Monday","Tuesday","Wednesday","Thursday","Friday"], "opens": "08:00", "closes": "17:00" }
  ],
  "areaServed": [ { "@type": "City", "name": "Springfield" }, { "@type": "City", "name": "Decatur" } ],
  "priceRange": "$$",
  "image": "https://acmeplumbing.com/og.png"
}

That is the canonical plumbing business as an entity. Answer engines can cite this.

What to do this quarter

For a small business in 2026, the practical move is:

1. Audit the website for Organization and (if applicable) LocalBusiness schema. 2. Add Service schema to every service page. 3. Convert the FAQ section, if there is one, into FAQPage JSON-LD with <details> elements. 4. Add AggregateRating if the business has a review source. 5. Validate every page that has JSON-LD using Google's Rich Results test.

That is the work. It is enough. For a small business running a static site with a schema-emitting plugin, the work is a single sprint and pays back continuously as the answer engines index and cite.