Schema Markup for AI Search: What Helps and What's a Myth

Schema markup for AI search is useful, but not in the way much of the advice suggests. Google says structured data isn’t required to appear in its generative AI features and that there’s no special schema.org markup for them, and no AI assistant’s documentation we know of says that JSON-LD earns you citations. What schema still does well is make pages eligible for rich results in Google Search and describe your organization and products in an unambiguous format.

This article separates what Google’s documentation says from the myths, summarizes two published studies, lists the schema types worth having, and shows how to check whether your markup changed anything in AI answers.

Key takeaways

• Google says structured data isn’t required for AI Overviews or AI Mode, and that there’s no special schema.org markup for generative AI search.
• Google still recommends structured data as part of SEO because it makes pages eligible for rich results.
• In a controlled Ahrefs study, pages that added JSON-LD saw no meaningful lift in AI citations compared with matched pages that didn’t.
• FAQ rich results are no longer shown in Google Search since May 2026, so FAQPage markup is not a route to extra visibility there.
• Worth having: Organization with sameAs, Article, Product or SoftwareApplication and BreadcrumbList, all matching what’s visible on the page.

What Google says: structured data isn’t required for AI features

Google’s guide to optimizing for generative AI features, the same one that calls AEO “still SEO” (see our AEO guide), is direct. Among the things you can ignore, next to llms.txt, it lists overfocusing on structured data: “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add.” It adds that structured data is still a good idea as part of your SEO strategy, because it helps with eligibility for rich results.

Google’s AI features documentation says the only requirement to appear in AI Overviews or AI Mode is that a page is indexed and eligible to be shown with a snippet. It also asks you to make sure your structured data matches the visible text on the page.

The introduction to structured data describes two jobs for markup: helping Google understand a page, and enabling richer search results. It recommends JSON-LD where possible. Required properties make a page eligible for a rich result, not guaranteed to get one.

Myths vs what the sources say

What the studies found, and why they seem to disagree

Ahrefs, May 2026 (correlation, then a controlled test). Ahrefs reports that AI-cited pages were almost three times more likely to have JSON-LD than non-cited pages. It then tracked 1,885 pages that added JSON-LD between August 2025 and March 2026 against 4,000 matched control pages, 30 days before and after. Relative to controls, citations moved by -4.6% in Google AI Overviews (statistically significant, unexplained by the authors), +2.4% in AI Mode and +2.2% in ChatGPT, the last two indistinguishable from zero. The authors’ own caveats: every page was already cited, schema types were pooled, and the window was 30 days.

Semrush, January 2026 (correlation). Semrush’s technical SEO study of URLs cited by ChatGPT Search and Google AI Mode found JSON-LD on about 40% of AI Mode-cited pages and about 30% of ChatGPT-cited pages, with Organization, Article and Breadcrumb most common. The authors state that it identifies correlations, not causation.

The two aren’t really in conflict. Well-maintained sites tend to add schema and also publish stronger content and earn more links, so schema shows up more on cited pages. When Ahrefs compared pages that added schema with similar pages that didn’t, the difference mostly disappeared. Neither study tells you what schema does for a page that isn’t cited yet.

Where schema still helps

• Rich results. For supported types, markup can make your listing more informative in Google Search. That affects how your result looks, which is separate from whether an AI feature uses your page.
• Entity clarity. Organization markup with sameAs links ties your site to your official profiles in a machine-readable way. Google says it uses structured data to understand page content.
• Product and organization facts. Pricing, offers and company details in markup restate what’s on the page in a consistent format.

The catch: assistants answer from text. If your pricing only exists in JSON-LD and not in the visible page, you’ve written it for the wrong reader. Put facts in plain text first, then mark them up.

The schema types worth having

Here is a minimal Organization example for a fictional CRM company. It goes in a <script type="application/ld+json"> tag on your homepage:

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Example CRM",
  "url": "https://www.example.com",
  "logo": "https://www.example.com/images/logo.png",
  "description": "CRM software for small B2B sales teams.",
  "sameAs": [
    "https://www.linkedin.com/company/example-crm",
    "https://x.com/examplecrm",
    "https://www.crunchbase.com/organization/example-crm"
  ]
}

Per schema.org’s Organization type, sameAs takes URLs of reference pages that unambiguously indicate the item’s identity. List profiles you actually control or that clearly describe you, not every directory you appear in.

Mistakes that waste the effort

• Marking up what users can’t see. Markup should describe content that’s visible on the page. Hidden or contradictory markup is a liability, not a shortcut.
• Conflicting blocks. Theme plugins and SEO plugins often both output Organization markup with different names or logos. Keep one source of truth.
• Invalid or incomplete markup. Check pages with Google’s Rich Results Test and the Schema Markup Validator after every template change.
• Adding FAQPage for the rich result. That result no longer shows in Google Search. Keep your FAQ section if it helps readers.
• Expecting citations. Treat schema as SEO hygiene, not as an AI visibility lever. For what tends to drive citations, see how AI assistants choose sources.

How to test schema’s effect on your AI visibility with Citegram

The evidence is mixed, so judge from your own prompts. A before-and-after comparison is an observation, not a controlled experiment: assistants, competitors and the web change at the same time. In the Ahrefs study, treated pages gained 43% in AI Mode citations on raw numbers, and control pages gained almost as much. For Google, also check Search Console, as covered in AI Overviews and your traffic.

1. Pick a fixed prompt set. Use the buyer questions where you’d expect your marked-up pages to be relevant.
2. Record a baseline. Citegram asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews, in background tabs, with the accounts already signed in to your browser. Run it several times.
3. Ship only the markup. Avoid content or design changes on the same pages during the test.
4. Re-run the same prompts from the same accounts after your pages have been recrawled.
5. Compare mentions, positions and cited sources. Citegram records whether your brand was mentioned, its position and the sources each answer cited, including which sources Google’s AI Overview cited. Reports show share of voice and the most cited domains over repeated runs.

Because answers come from your own accounts, personalization can affect them. Keep the setup identical between rounds.

FAQ

Does FAQ schema help with AI Overviews?

Google says no special markup is needed for AI Overviews, and FAQ rich results stopped appearing in Google Search on May 7, 2026. A clear FAQ section can still help readers, and it’s text any system can read. The markup itself isn’t a documented route into AI answers.

Does ChatGPT read JSON-LD?

OpenAI doesn’t document whether ChatGPT uses structured data when it selects or cites sources. Its crawlers can fetch your HTML, which includes JSON-LD, but no published study shows that markup causes ChatGPT citations. Put the facts that matter in visible text.

What schema should a SaaS website have?

Start with Organization on the homepage (with sameAs links to your official profiles), Article on blog posts, SoftwareApplication or Product on product and pricing pages, and BreadcrumbList site-wide. Make sure every marked-up value also appears on the page.

Is schema a ranking factor?

Google’s documentation presents structured data as a way to understand content and become eligible for rich results, not as a ranking boost. Rich results can change how your listing looks and how often it’s clicked, which is different from where it ranks.

Measure before and after

If you ship schema changes, measure the answers rather than guessing. The free plan tracks one prompt across all major assistants, forever, and Pro adds unlimited prompts for $10 a month. See pricing.