Schema markup is the most reliable lever in GEO. The work is mechanical, the gains are real, and most sites are still doing it wrong or not at all. This piece is the field-tested shortlist of what to deploy and how — based on what we actually see move citations.

Why schema matters more for AI than for Google

Traditional search engines use schema to enable rich results (star ratings, recipe cards, FAQ accordions in SERPs). Useful, but not strictly required to rank.

AI engines use schema as a primary input to their entity model. When an AI assistant decides whether to cite you, it's asking: is this site a credible source about [topic]? Schema lets you answer that question explicitly — "we're a software company called X, we serve Y type of customer, we have these products, these are our prices, this is our address, here are our reviews." Without schema, the engine has to infer all of that from prose, and it often gets it wrong.

The required minimum (every page)

Organization (or LocalBusiness)

Put this on every page. Tells AI engines who publishes the site. Critical fields:

  • name — your business name, exactly as you want it cited
  • url — your canonical homepage URL
  • logo — full URL to a square logo, ideally 512×512 minimum
  • description — one sentence on what you do (category, target customer, outcome)
  • sameAs — array of your social/profile URLs (LinkedIn, Crunchbase, etc.)
  • contactPoint — email and/or phone, with contactType

If you're a local business (physical location, geographic service area), use LocalBusiness instead, which inherits from Organization and adds address, geo, openingHours, priceRange, and area-served fields.

WebSite

Companion to Organization. Tells engines this is a website and what its name is. Include the optional potentialAction with a SearchAction if you have site search — it enables AI engines to suggest searches on your site.

BreadcrumbList

On internal pages, helps AI engines understand your site's hierarchy. Especially useful if you have nested content (blog → category → article, or shop → category → product).

The high-leverage page-type schemas

FAQPage — the single highest-impact schema for citations

If you have Q&A content anywhere on your site, wrap it in FAQPage schema. AI engines treat this as gold because the format matches exactly what they're trying to extract: a question, followed by an answer. We consistently see FAQPage schema correlate with citation rate gains across nearly every category.

Don't fake it. The questions in your schema must match real Q&A on the page. Mismatched schema gets ignored or penalized.

Article

Every blog post should have Article schema (or BlogPosting, a subtype). Critical fields: headline, author (with name and ideally URL), datePublished, dateModified, image, publisher. The dates matter — AI engines weight recency heavily, and a stale article without a recent dateModified ranks below comparable updated content.

Product (with Offer and AggregateRating)

For e-commerce. The basics aren't enough — include the full Offer block with price, priceCurrency, availability, and itemCondition. If you have honest reviews, include AggregateRating. Don't fabricate ratings — engines now cross-check this against external review platforms.

Service

For service businesses. Pair with Offer for pricing. Include serviceType (use a recognized category), provider (your Organization), areaServed, and availableChannel.

HowTo

For step-by-step content. Each step gets its own HowToStep entry with text, image, and optional tools/supplies. AI engines pull these for "how do I..." queries with remarkable consistency.

Review and AggregateRating

Honest reviews of products, services, or businesses. Cross-checked against external platforms. Faked reviews actively hurt your AI credibility.

Schemas that are less critical than people think

Some types get a lot of attention but produce minimal AI citation gains in 2026:

  • Event — useful for events specifically, not broadly applicable
  • Recipe — recipe-vertical only; massive in food, irrelevant elsewhere
  • Video — helpful if you have video, but the gains are mostly in Google Video, not AI citations
  • Speakable — built for voice search, has narrow current relevance
  • Course — relevant only for education businesses

Deploy these if they fit your content type. Don't agonize about them if they don't.

How to deploy correctly

Three rules:

1. Use JSON-LD, not Microdata or RDFa. JSON-LD is what Google recommends, what AI crawlers parse most reliably, and what tooling validates best. Microdata and RDFa work but are increasingly second-class.

2. Put schema in the page's HTML head, not injected via JavaScript. Same reason as content: AI crawlers often don't execute JS, so JS-injected schema gets missed. Use server-side rendering or static generation.

3. Validate everything. Bad schema is often worse than no schema. Use Google's Rich Results Test and Schema.org's Schema Markup Validator. Fix every error and warning before deploying.

A common-mistake checklist

  • Mismatched schema and visible content — your FAQPage schema says one question but the page shows a different one. Engines detect this and either ignore the schema or penalize.
  • Stale dateModified — your schema says the article was updated in 2024 but you've changed it three times since. Update the field every time you change the content.
  • Missing or generic description — Organization schema with description "We are a leading provider of solutions" is functionally useless. Be specific.
  • Inconsistent NAP — your LocalBusiness schema says one address but Google Business Profile says another. This breaks the entity model AI engines build of you.
  • Fake reviews — AggregateRating with no underlying review platform data is a red flag. Engines cross-check.

How long it takes to see effects

Schema deployment shows up in Google's rich results within days to weeks. AI engine citation gains take longer — usually 4–8 weeks as crawlers re-index your site, the engines re-train or update their search index, and the structured data filters into the underlying entity models. Don't panic if you deploy schema and don't see immediate citation gains. The compounding is slow but durable.

What Reffed does on the schema layer

When you connect a site to Reffed, the platform audits your existing schema, deploys missing types, fixes errors, and keeps everything current as your content changes. We also handle the parts most plugin-based schema tools skip — like keeping dateModified accurate, ensuring multi-page consistency, and matching schema to actual visible content.

Free schema audit

Reffed checks every schema type on your site, validates them, and identifies what's missing. Run it in 60 seconds.

Related reading