Skip to content
7.11Advanced9 min

Schema Markup for AI Search Engines: Confirmed and Measurable

Lucas Blochberger··Updated 11 June 2026
Definition

Schema markup has been officially confirmed AI infrastructure since March 2025. Google, Microsoft and ChatGPT confirm the use of structured data for their generative features. AirOps found that pages with 3+ schema types paired with clean heading hierarchy show 2.8x higher AI citation rates.

Key Takeaways

  • ✓Google and Microsoft confirmed the use of Schema for AI features in March 2025
  • ✓ChatGPT confirms Schema usage for product results
  • ✓Pages with 3+ Schema types achieve 2.8x higher AI citation rates
  • ✓LLMs with Knowledge Graph structure: 54.2% vs 16.7% accuracy (3.25x)
  • ✓FAQPage Schema: Pages with it are 3.2x more likely in AI Overviews
  • ✓NLWeb (Microsoft's initiative): Schema as MCP server for AI agents
  • ✓Counterpoint: Only 1.8% of AI-cited sources use FAQPage Schema

Schema Markup has crossed the threshold from optional to essential — confirmed by three of the world's leading tech companies.

The triple confirmation

In March 2025, Google publicly confirmed: Structured data is critical for modern search features. Microsoft confirmed at SMX Munich that Schema Markup helps their LLMs understand content for Copilot. ChatGPT confirmed its use for product results.

Measurable impact

AirOps analyzed 12,000+ URLs and found that pages with 3 or more Schema types combined with clean heading hierarchy have 2.8x higher AI citation rates. The data.world benchmark showed that LLMs with knowledge graph structure achieve 54.2 percent accuracy versus 16.7 percent without — a 3.25x improvement.

Schema type priority

FAQPagehas the strongest demonstrable effect: pages with it are 3.2x more likely to appear in AI Overviews.Articlewith Author, datePublished and dateModified signals timeliness and authority.Organizationwith sameAs links to Wikidata and Wikipedia enables entity verification.Productwith Offer and AggregateRating is essential for commerce-related AI queries.Personwith knowsAbout supports expert identification.

The important counterpoint

AccuraCast found that among AI-cited sources, only 1.8 percent use FAQPage Schema, and Reddit — with zero Schema — remains one of the most cited sources overall. Schema functions as an accelerator for borderline content, not a prerequisite for citation.

NLWeb: Schema as AI infrastructure

Microsoft's NLWeb initiative (May 2025), created by Schema.org founder R.V. Guha, elevates Schema from the annotation to the application layer. Each NLWeb instance functions as an MCP server (Model Context Protocol), enabling AI agents to discover and query structured content directly.

Data & Statistics

Strukturierte Daten sind laut Google nicht erforderlich für generative KI-Suche (kein spezielles schema.org-Markup nötig), werden aber als Teil der SEO-Strategie für Rich Results empfohlen.

Google Search Central - AI Features Optimization Guide [international] (2026)

1.885 Seiten mit neu hinzugefügtem Schema vs. 4.000 Kontrollseiten (Difference-in-Differences): Google AI Overviews minus 4,6 %, AI Mode plus 2,4 %, ChatGPT plus 2,2 % - kein signifikanter Uplift auf einer Plattform.

Ahrefs Blog - 'We Tracked 1,885 Pages Adding Schema' [international] (2026)

Nach Schema-Implementierung (319 Prompts, US): AI Overviews plus 611 %, AI Mode plus 42 %, ChatGPT minus 71 %, Copilot minus 64 %, Gemini minus 35 %, Perplexity 0 %; nur 1 von 7 Plattformen (Gemini) konnte rohes JSON-LD auslesen.

Otterly.AI Blog - 'Schema Markup's Real Impact on AI Search' [international/US] (2026)

GEO kann die Sichtbarkeit einer Quelle in generativen Antworten um bis zu 40 % steigern; die stärksten Hebel sind inhaltsbasiert (Quellen, Statistiken, Zitate, Autorität) statt technisches Markup.

Princeton University et al. - 'GEO: Generative Engine Optimization' (KDD 2024), Aggarwal et al. [international] (2024)

JSON-LD-Verbreitung wuchs von 34 % (2022) auf 41 % (2024) der gecrawlten Seiten (schnellst wachsendes Format, von Google bevorzugt); häufigste Typen: WebSite 12,73 %, Organization 7,16 %, BreadcrumbList 5,66 %, LocalBusiness 3,97 %.

Web Almanac 2024 (HTTP Archive) - Structured Data [international] (2024)

20 % der österreichischen Unternehmen (ab 10 Beschäftigten) nutzten 2024 KI-Technologien (2023: knapp 11 %); Information und Kommunikation 61 %; Texterkennung/-verarbeitung 65 % und Sprachgenerierung 41 % der KI-nutzenden Unternehmen.

Statistik Austria - IKT-Einsatz in Unternehmen 2024 (PM 13 449-215/24) [Österreich] (2024)

AI-Overviews-Auslöserate stieg von 6,49 % (Jan 2025) auf 24,61 % (Jul 2025) und pendelte sich bei 15,69 % (Nov 2025) ein; Zero-Click-Rate auf identischen Keywords sank von 33,75 % auf 31,53 %; über 10 Mio. Keywords analysiert.

Semrush - AI Overviews Study [international] (2025)

ChatGPT-E-Commerce-Traffic konvertierte mit 1,81 % um 31 % höher als nicht-markenbezogener Organic-Traffic (1,39 %); 94 Sites, 9,46 Mio. Organic- vs. 135.000 ChatGPT-Sessions, 12 Monate GA4 (2025).

Search Engine Land (Visibility Labs study) [international] (2025)

Microsoft Bing (Fabrice Canel) bestätigte auf der SMX Munich (März 2025), dass Schema-Markup den LLMs von Microsoft (Copilot/Bing) hilft, Inhalte zu verstehen; Empfehlung, frische Inhalte über IndexNow zu pushen.

Search Engine Land (Barry Schwartz) - 'Microsoft Bing/Copilot use schema for its LLMs' [international] (2025)

FAQ

Do AI search engines like ChatGPT and Perplexity read Schema Markup?
Predominantly not. In a fetch test by Otterly.AI, 6 out of 7 AI platforms could not retrieve or correctly interpret Schema Markup when directly queried; only Gemini delivered the correct JSON-LD. ChatGPT, Perplexity, Claude, and Copilot rely predominantly on visible, rendered page content. Google systems (AI Overviews, AI Mode, Gemini) come closest to Schema, as they build on Google's index. Microsoft has also officially confirmed via Fabrice Canel that Schema helps its LLMs with understanding.
Does Schema Markup generate more AI citations?
The evidence is contradictory and shows no reliable uplift. Ahrefs tracked 1,885 pages that added Schema and found no significant increase in citations on any platform (Google AI Overviews even showed minus 4.6 percent). Otterly, however, reported plus 611 percent for AI Overviews, but simultaneously minus 71 percent for ChatGPT. Both study providers conclude that Schema is an SEO lever, not a GEO growth lever. Schema should be deployed as a foundation, not as a citation booster.
Is Schema Markup necessary for AI search at all?
According to Google, it is not required. Google states in its AI Optimization Guide that structured data is not required for generative AI search and no special schema.org markup needs to be added, but it remains sensible to use it as part of the overall SEO strategy, as it supports eligibility for Rich Results. Schema is therefore not a mandatory factor for AI visibility, but an established SEO foundation in the Google and Microsoft ecosystem.
What is more important for AI visibility: Schema or Content?
Content. The Princeton GEO study shows that GEO can increase visibility in generative answers by up to 40 percent, with the strongest levers being content-based: citing sources, incorporating statistics, using quotes, as well as authority and linguistic clarity. Technical markup is not among these strongest levers. The correct order is therefore: first verifiable, clearly structured content with key statements upfront, then Schema as an amplifier that unambiguously marks up this content.
Which Schema types are relevant for AI visibility?
Primarily Organization (defines the company as an entity), Person and Author (support E-E-A-T and authorship), sameAs (links the entity with authoritative references like Wikidata or LinkedIn), FAQPage (structures questions and answers), Article (identifies expert content), and Product (basis for Agentic Commerce with price and availability data). What matters is not the number of types, but their correctness and consistency with the visible content.
Should one focus on Schema Markup or llms.txt?
On Schema. Structured data is widely adopted and part of the existing crawling ecosystem: JSON-LD grew to 41 percent of crawled pages (2024) according to Web Almanac and is the fastest-growing format. llms.txt, on the other hand, is largely ignored by most AI crawlers so far. Those who invest in llms.txt and neglect Schema are prioritizing the wrong way around. Schema is the more robust and crawler-readable lever.
How do you measure the effect of Schema Markup on AI search?
On three levels and in the right place. First, Schema validation with Google's Rich Results Test and the Schema.org Validator. Second, AI referral traffic separately in GA4 via referral sources and custom channel groups to isolate traffic and conversions from ChatGPT, Perplexity, or Copilot. Third, brand mention and citation monitoring with tools like Otterly, Ahrefs Brand Radar, or Semrush. Important: Schema effects show primarily in Google Rich Results and AI Overviews, not in ChatGPT citations.

Related Articles

How does your website perform?

Get a free, AI-powered SEO report of your website by email – technical SEO, on-page, keywords & competitors. No obligation.

Get a free SEO audit →
Previous← DACH GEO Strategy: English Language Bias and Market OpportunitiesNextStatistics and Data for GEO: The 41% Effect →