Structured Data for GEO: The 6 Schema Types AI Assistants Actually Use in 2026

The schema.org types AI assistants like ChatGPT and Perplexity actually use in 2026 are Organization, WebSite, FAQPage, BlogPosting (or Article), Product, and HowTo — in roughly that order of impact. Everything else is either redundant, niche, or so weakly signaled that adding it changes nothing about how your business shows up in AI answers.

This post is the practical version of "what should I actually mark up on my site." Not the full schema.org catalog of 800+ types, not a theoretical exercise, just the six types that move the needle in 2026 when buyers are asking AI assistants for recommendations instead of typing into Google.

Why Structured Data Matters MORE for AI Than for Google

Here is the part most SEO articles get wrong. Structured data was already useful for Google. For AI assistants, it is fundamentally more important, and the reason is mechanical, not philosophical.

Google ranks pages. The classic search algorithm reads your visible content, ignores some of it, infers facts, ranks the page against competitors, and shows a list. Schema markup helps, but Google has been doing semantic inference at scale for so long that a well-structured page without schema still ranks. The crawler can figure out that your H1 is a product name, your $49 nearby is a price, and your star rating image means reviews exist. Schema makes that confirmation faster, but Google does not strictly need it.

AI assistants extract facts. When ChatGPT, Perplexity, or Claude answer "what is the best CRM for solo agents under $50 a month," the system is not ranking your page against ten others — it is extracting specific data points (name, price, rating, category) and stitching them into a sentence. If those data points are sitting in a JSON-LD block, the AI grabs them directly with near-zero ambiguity. If they are buried in marketing prose, the AI may infer them, may infer them wrong, or may simply skip your site for a competitor whose data is easier to parse.

Structured data is pre-chewed food for AI assistants. Google can eat the steak. ChatGPT prefers the cube.

This is the part that changes the calculation in 2026. If you were on the fence about schema in 2023 because "Google figures it out anyway," that excuse is gone. The AI traffic share is growing, and AI traffic depends much more directly on whether your facts are explicit and machine-readable. We covered the broader shift in our piece on what is GEO (generative engine optimization) — structured data is one of the three or four levers that actually move GEO performance.

The 6 Schema Types That Carry the Most Weight in 2026

Not all schema is created equal. After auditing thousands of customer sites through EMAX Studio's Quick Scan, six types come up over and over as the ones AI assistants actually surface. Get these six right and your site is in the top 5 percent of GEO-ready sites in your industry. Add a seventh and you are gold-plating.

1. Organization — Who You Are

Organization schema tells AI assistants what your business is, who runs it, when it started, and where to find it elsewhere on the web. It is the single most important schema type for GEO because almost every AI answer that mentions a company pulls from this object.

Fields that matter: name, legalName, url, logo, sameAs (your social profiles and external mentions), founder, foundingDate, description, email. The sameAs array is the killer field — it tells the AI "this is the same business as that LinkedIn profile, that X account, that Crunchbase page." Without it, the AI cannot confidently merge your identity across sources.

Place Organization schema on your homepage and ideally on your About page. Do not duplicate it across every page — one canonical Organization block on the root domain is plenty.

2. FAQPage — Direct Q&A That AI Loves to Quote

FAQ markup is the schema type AI assistants quote most directly. When a user asks Perplexity "does company X offer refunds," the AI looks for an FAQ that contains exactly that question and pulls the answer verbatim. We covered this in detail in FAQ schema for AI assistants, but the short version is this: every page on your site that has a real FAQ section should have FAQPage schema.

Required fields: mainEntity array with Question objects, each containing name (the question) and acceptedAnswer with Answer.text (the answer). Keep questions phrased the way users actually ask them. Keep answers under 80 words.

3. BlogPosting / Article — Authority and Authorship

Every blog post and editorial article needs BlogPosting (or its parent Article) schema. This tells the AI when the content was published, who wrote it, what the headline is, and where the canonical image lives. AI assistants weight recent, authored content more heavily than anonymous undated content — and the only way they reliably know your post is recent and authored is via schema.

Critical fields: headline, author (as a Person object with name and ideally URL), datePublished, dateModified, image, publisher (linking back to your Organization schema). The dateModified field is underrated — it lets you signal that you have refreshed an older post, which AI assistants increasingly favor for time-sensitive queries.

4. Product — Names, Prices, Ratings

If you sell anything online, Product schema is non-negotiable. AI assistants use it to answer comparison queries ("show me a meditation app under $10 a month with at least 4 stars") and recommendation queries ("recommend a coffee subscription"). Without Product schema, your offers exist only as marketing prose, which the AI may or may not parse correctly.

Required fields: name, description, image, brand, offers (with price, priceCurrency, availability), and where applicable aggregateRating (with ratingValue and reviewCount) and review arrays. The ratings are the part that wins comparison queries — and the part most sites forget to mark up.

5. HowTo — Step-by-Step Procedures

HowTo schema is for any page that walks through a process — "how to set up X," "how to assemble Y," "how to file Z." AI assistants love HowTo content because it answers procedural queries directly, and the structured step array makes extraction trivial.

Fields: name, description, step array (each step a HowToStep with name, text, optionally image). If your business has support documentation, tutorials, or onboarding content, HowTo schema makes it surface in answers like "how do I [verb] with [your product]." That is high-intent traffic.

6. WebSite — Site-Wide Identity and Search

WebSite schema sits at the root of your domain and tells AI assistants the name of your site (which can differ from your Organization name), your primary URL, and optionally a potentialAction defining your internal search endpoint. The internal search hookup powers Google's sitelinks search box, but more importantly it gives AI assistants a clean way to reference "search inside this site for X."

It is a small, quiet schema, but pairing WebSite with Organization on your homepage is the cheapest two-block upgrade you can make. Five minutes of work, lifelong payoff.

How to Add Schema in 10 Minutes

Adding structured data is not a project. It is a five-step micro-task you can finish on a coffee break.

Step 1: Choose your top 3 page types. For most businesses these are: homepage, blog post template, product/service page. If you sell, swap "service page" for "product page." If you are a content site, swap it for "category page."

Step 2: Pick the matching schema. Homepage gets Organization + WebSite + FAQPage (if you have a home FAQ). Blog post template gets BlogPosting. Product page gets Product. That is it — three templates cover 90 percent of pages on a typical site.

Step 3: Add JSON-LD inside the <head> tag. Use a <script type="application/ld+json"> block. Do not put schema in the body. Do not use Microdata or RDFa (more on that below).

Step 4: Validate with Google's Rich Results Test. Paste your URL into search.google.com/test/rich-results. It tells you which schema types it detected, which fields are missing, and any errors. Free, instant, no signup.

Step 5: Monitor in Google Search Console. Under "Enhancements" you see which schema types Google has indexed and any errors. Watch this weekly for the first month, then monthly.

Total time for a site with 3 templates: about 10 minutes once you have the templates ready. Less if you use a CMS plugin (Yoast, RankMath, or built-in WordPress block schema).

JSON-LD vs Microdata vs RDFa

Three syntaxes exist for structured data. You should only use one.

JSON-LD is the modern, Google-preferred format. It sits in a separate <script> block in the head, completely separate from your visible HTML. Easy to read, easy to maintain, easy to template, no risk of breaking your layout. Every example in this post is JSON-LD.

Microdata embeds schema directly into HTML attributes (itemscope, itemtype, itemprop). It works, but it clutters your markup, breaks easily during redesigns, and is harder to debug. Use only if your CMS forces you.

RDFa is similar to Microdata but with different attribute syntax. It is academic, beloved by semantic-web purists, and used by almost nobody in commercial SEO. Skip it.

Google's published guidance prefers JSON-LD, the AI assistants we have tested parse it most reliably, and it is by far the easiest to maintain. There is no scenario in 2026 where Microdata or RDFa is the right answer for a new implementation.

Schema Type Comparison: What to Use Where

If you only have time for two, do Organization and FAQPage. If you have time for four, add BlogPosting and Product. The other two are upgrades, not foundations.

A Worked Example: Three Schema Blocks for a Real Site

Here is what the schema setup looks like for a small SaaS company called Acme Studio with a hosted blog and a product page. Three JSON-LD blocks, drop-in templates.

Organization (on every page, in the <head>)

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Acme Studio",
  "legalName": "Acme Studio Inc.",
  "url": "https://acmestudio.com",
  "logo": "https://acmestudio.com/logo.png",
  "description": "AI-powered design tool for small teams.",
  "foundingDate": "2024-03-15",
  "founder": {
    "@type": "Person",
    "name": "Jane Smith"
  },
  "email": "hello@acmestudio.com",
  "sameAs": [
    "https://www.linkedin.com/company/acmestudio",
    "https://twitter.com/acmestudio",
    "https://www.crunchbase.com/organization/acmestudio"
  ]
}
</script>

WebSite (on homepage only)

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "WebSite",
  "name": "Acme Studio",
  "url": "https://acmestudio.com",
  "potentialAction": {
    "@type": "SearchAction",
    "target": "https://acmestudio.com/search?q={search_term_string}",
    "query-input": "required name=search_term_string"
  }
}
</script>

BlogPosting (on every blog post)

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "BlogPosting",
  "headline": "How to Set Up Acme Studio in 5 Minutes",
  "image": "https://acmestudio.com/blog/setup-guide/hero.jpg",
  "datePublished": "2026-06-01",
  "dateModified": "2026-06-09",
  "author": {
    "@type": "Person",
    "name": "Jane Smith",
    "url": "https://acmestudio.com/team/jane-smith"
  },
  "publisher": {
    "@type": "Organization",
    "name": "Acme Studio",
    "logo": {
      "@type": "ImageObject",
      "url": "https://acmestudio.com/logo.png"
    }
  },
  "description": "A step-by-step guide to setting up Acme Studio for your first project."
}
</script>

That is roughly 40 lines of JSON. It takes longer to read than to ship. Once these three templates are in your codebase, every new page and post inherits structured data automatically.

Common Mistakes That Kill Your Schema

Most schema problems are not exotic. They are the same five mistakes, repeated across thousands of sites.

Forgetting required fields. Google's Rich Results Test will flag these. The most common omission is publisher on BlogPosting (required) and image on Product (required for rich results).

Mismatching schema and visible content. If your Product schema says price is $49 and the visible page says $79, you are signaling spam. Google penalizes this. AI assistants distrust the data and may skip your site entirely. Schema must match what users see.

Duplicate Organization across pages. One canonical Organization on the homepage, optionally referenced via @id on internal pages. Do not put a fresh Organization block on every single page with slightly different data — it confuses crawlers and AI parsers.

Broken sameAs links. The whole point of sameAs is identity stitching. If your LinkedIn URL 404s, the link is worthless and you lose the trust signal. Audit sameAs URLs quarterly.

Invalid date formats. Schema.org expects ISO 8601 dates (2026-06-09T14:30:00+00:00 or just 2026-06-09). Anything else gets silently ignored. The number of sites that ship "datePublished": "June 9, 2026" and wonder why nothing works is embarrassing.

Never validating. Schema fails silently. A typo in a property name (founderr instead of founder) just drops that field with no warning. Run every template through the Rich Results Test before deploying, and again after any redesign. Five minutes. Saves months of wondering why your GEO score has not moved.

Frequently Asked Questions

Do I need all 6 schema types?

No. Most sites need three: Organization (homepage), FAQPage (where FAQs exist), and BlogPosting (every post). Product comes in for ecommerce. HowTo for tutorial-heavy sites. WebSite is a five-minute add on the homepage. Pick what matches your actual content, not what looks impressive.

How do I know if AI assistants actually use my schema?

The honest answer is that AI providers do not publish detailed analytics on which schema they parsed. What you can do: ask the same question to Perplexity or ChatGPT before and after adding schema, and watch whether your site gets cited or whether your specific facts (price, rating, founding date) show up correctly in the AI's answer. Anecdotally, adding Organization + FAQPage to a site that had neither produces visible changes in AI citations within 2 to 6 weeks as the AI assistants re-crawl.

Can I overdo it with schema?

You can overdo it in two specific ways. First, by stuffing schema with facts that do not appear on the visible page — Google flags this as deceptive markup. Second, by adding so many irrelevant schema types (Recipe schema on a SaaS homepage, Event schema on a static blog) that the picture gets noisy. Stick to schema that maps to what is genuinely on the page. More is not better.

What about LocalBusiness schema for brick-and-mortar shops?

LocalBusiness (and its many subtypes — Restaurant, Dentist, Bakery) is a critical addition for any business with a physical location. If you serve walk-in customers, replace the Organization block with the relevant LocalBusiness subtype. The fields are similar, plus address, geo (lat/long), openingHours, and priceRange. AI assistants use this heavily for "near me" and local recommendation queries.

Does schema also affect traditional SEO ranking?

Indirectly, yes. Google has stated that schema is not a direct ranking factor, but rich results (review stars, FAQ accordions, product cards) increase click-through rate, and click-through is correlated with ranking. So schema does not push you up the rankings directly, but the rich snippets it enables can win you clicks against higher-ranked competitors. We dug into this in AI SEO vs traditional SEO — schema is one of the few tactics that scores well in both worlds.

How long until I see results from adding schema?

For Google rich results, expect 2 to 6 weeks for Search Console to pick up the markup and start showing it in search. For AI assistants, citations can update within days for high-traffic sites that Perplexity and ChatGPT crawl frequently, or up to 2 months for lower-traffic sites that get re-crawled less often. There is no instant gratification with schema. There is steady, compounding payoff.

The Honest Bottom Line

Structured data is not glamorous. It does not produce a visible-on-the-page wow moment. Nobody buys your product because your schema is well-formed. But in 2026, with AI assistants increasingly being the layer between you and your customers, schema is the difference between being cited as a fact and being skipped as too much work to parse.

The six schema types above are not optional anymore for serious businesses. They are basic hygiene, like having SSL or a mobile-responsive layout. The companies that ship them get cited. The companies that do not get replaced by competitors in answers users never even saw a search results page for.

The good news: it takes 10 minutes per template once you have the patterns down. If you want a free check on which schema types are missing from your site right now — and where your GEO score is overall — emax.studio runs a free 90-second Quick Scan that audits structured data across 5+ schema types as part of its GEO sub-score. No signup needed to see the result.

Schema is the cheapest GEO investment you will ever make. Make it this week.

Follow EMAX Studio: Instagram | YouTube | Facebook