Schema Markup On-Page: Required for Rich Results, Not a Booster for AI Citations

Why schema markup on-page matters for B2B in Austria

Schema markup, also called structured data, translates a web page's content into a machine-readable vocabulary. Search engines, and increasingly AI systems, use it to recognize explicitly what a page describes: a company, an article, a service, a person. In the DACH region, this understanding is gaining weight because search behavior is shifting.

The numbers from Austria document the change. According to the AI Readiness Study 2025 by the Handelsverband, Austria's retail trade association, 68 percent of the Austrian businesses surveyed use AI, 37 percent of them as an intelligent search engine and 49 percent for writing text. Nearly three quarters of the Austrian population have heard of ChatGPT, and one in five uses the service several times a week. If you want to be found and correctly understood in this environment, you need a clean machine-readable foundation.

At the same time, classic search is changing measurably. After Google launched AI Overviews in Germany and Austria in March 2025, a study conducted by Wordsmattr.io found that clicks fell by an average of 17.8 percent and CTR by 14 percent, while impressions stayed nearly stable, declining by only 1.2 percent. The sample covered 125 optimized landing pages and more than 1,000 keywords in Germany and Austria. The message for B2B decision-makers is clear: visibility increasingly happens in the answer itself, not only on the landing page. Schema markup is one building block for being represented cleanly in this answer layer.

What schema markup on-page really does, and what it doesn't

Honesty matters more here than any promise. Structured data has a clearly documented purpose and a clearly documented limit.

The documented benefit is rich result eligibility. Google is explicit: pages that are missing required properties are not eligible for rich results. Without correct markup, there are no star ratings and no other enhanced results in the SERP. Schema is the technical prerequisite for Google to offer these display formats at all.

The often-claimed effect as a direct booster for AI citations, on the other hand, does not hold up. A 2026 Ahrefs analysis tracked 1,885 pages that introduced JSON-LD schema between August 2025 and March 2026, measured against 4,000 control pages in a difference-in-differences design. The result is sobering: Google AI Overviews minus 4.6 percent, AI Mode plus 2.4 percent, ChatGPT plus 2.2 percent, and the last two values are statistically indistinguishable from zero. The study's conclusion: schema produced no meaningful increase in citations on any platform.

The honest conclusion for DACH B2B is therefore: schema markup is not a lever that directly forces AI citations. It is a foundation. It makes content unambiguously interpretable, secures eligibility for enhanced displays, and reduces ambiguity about entities. Anyone who sells schema with the promise of a guaranteed increase in citations is overstretching the evidence. Anyone who implements it as a clean machine-readable foundation is doing what the evidence supports.

Implementing JSON-LD correctly

JSON-LD has become the standard. According to the Web Almanac 2024, 41 percent of pages use JSON-LD, up from 34 percent in 2022. It is Google's preferred format and the fastest growing of the three options. JSON-LD keeps structured data separate from the visible HTML, which makes it easier to maintain than Microdata or RDFa, both of which are woven into the markup.

A few points matter most during implementation:

• Format and placement: JSON-LD as a script block in the <head> or at the end of the <body>. There is no need to weave it inline into the visible HTML.
• Link entities via @graph: Instead of isolated snippets, combine the objects on a page (Organization, WebSite, Article, Author) in one @graph and reference them via @id. This creates a connected picture instead of disjointed fragments.
• Truthfulness: Google requires that structured data be a true representation of the page content and that no content be marked up that is not visible to readers. Markup without a visible counterpart is a policy violation.
• Complete required fields: Because missing required properties rule out rich result eligibility, every required property belongs in the markup.
• Validate: Check all markup with the Rich Results Test and the Schema.org Validator before it goes live. The first checks eligibility with Google, the second general syntax conformance.

Which schema types to implement first for B2B in the DACH region

Not every type delivers the same value. For Austrian B2B providers, this order makes sense:

• Organization: Defines the company as an entity with name, logo, contact details, and sameAs references to LinkedIn and Wikidata. This linking creates entity clarity and helps machines attribute the brand unambiguously.
• WebSite: Represents the domain as a whole and can mark up the internal site search.
• BreadcrumbList: Mirrors the navigation hierarchy. Since January 2025, Google shows the breadcrumb path in search results only on desktop.
• Article: For expert articles and blog posts, including author, datePublished, and dateModified as E-E-A-T signals.
• LocalBusiness: For locations with address, opening hours, and geographic reference, relevant for regional visibility in Austria.
• Service or Product: For the specific service or product offering that describes the business.

The common denominator is the sameAs link to authoritative profiles such as Wikidata and LinkedIn. It anchors your own entities in a larger knowledge graph and reduces the risk of confusion, especially for generic company names in German-speaking markets.

Schema in combination with the rest of your on-page architecture

Structured data does not replace a clean on-page foundation; it builds on it. Markup that describes content poorly structured in the visible HTML has no effect or even violates the guidelines.

The foundation consists of several layers. Title tags and meta descriptions state the promise that appears in the SERP and in the answer layer. A logical H1-to-H6 hierarchy structures the content so that both readers and parsers grasp how it is organized. Semantic HTML with the correct elements for articles, navigation, and secondary content provides the structural meaning that JSON-LD then makes explicit. Schema and HTML have to tell the same story, because Google's truthfulness requirement applies throughout. If you keep your on-page basics clean, you create the conditions for structured data to make any contribution at all.

GEO on-page beyond schema

If schema doesn't directly increase AI citations, the question is what works instead. The answer lies in Generative Engine Optimization, or GEO for short, which works at the content level. The scientific basis comes from the paper by Aggarwal et al., presented at KDD 2024. It shows that GEO methods can increase visibility in generative engine responses by up to 40 percent.

An important caveat: this study is international and academic, not DACH-specific evidence from practice. The documented core finding is the visibility gain from GEO as a whole. The individual tactics associated with GEO in the broader discussion concern content quality: citable statistics with a clear source, embedded quotations, unambiguous source attributions, and high entity density, meaning precise and frequent use of relevant technical terms and proper names. For B2B content in the DACH region, this means in practice: back up claims, name sources, use technical terms consistently. This substance has a stronger effect in the answer layer than pure markup, which only describes the packaging.

Preparing for featured snippets and position 0

Featured snippets are a bridge between the classic SERP and the answer layer. You can prepare for them deliberately, but their effect on traffic cuts both ways.

Structurally, question-and-answer blocks, tables, lists, and compact definitions of roughly 40 to 60 words help. These formats give Google extractable answer components. Clean semantic HTML and clear headings increase the chance of being shown as a snippet.

Click cannibalization is part of the honest picture, though. According to a 2017 Ahrefs study of two million featured snippets, around 12.29 percent of US search queries had a featured snippet. When a snippet ranked in position 1, it received on average only about 8.6 percent of clicks. The organic result directly below dropped from around 26 percent of clicks without a snippet to around 19.6 percent with a snippet. These figures date from before AI Overviews. The lesson: a featured snippet secures visibility and brand presence, but does not guarantee more clicks. Combined with the click decline caused by AI Overviews, value shifts from click counts to presence in the answer.

Deprecation management and future-proofing

Schema is not a static field. Google is actively reducing the number of markup types that generate visible rich results. Google has not shown HowTo rich results since September 2023, and has not shown FAQ rich results since May 7, 2026. This significantly shifts the ROI of individual types.

The strategic consequence is a clear separation. Some markup types still deliver a visible SERP benefit, such as Product, Organization, or Breadcrumb (the latter now only on desktop). Other markup types no longer generate a rich result but can still be useful because they help machines and LLMs understand the content. Since May 2026, FAQ markup belongs entirely in this second category: no display advantage anymore, but possibly still a contribution to machine interpretation. B2B teams should prioritize their markup portfolio by whether a type still brings a visible benefit or only serves machine understanding. Effort goes first into the types with demonstrable eligibility.

Operationalizing E-E-A-T on-page

Trust signals can be expressed in structured form. Google and LLMs can then trace who wrote a piece of content and when it was updated.

Several elements can be put into practice. Author and Person schema name the responsible person and can point to their professional profiles via sameAs. Visible author boxes give the markup a visible counterpart, which satisfies Google's truthfulness requirement. Publisher information in the Organization object identifies the publishing entity. Date markup with datePublished and dateModified signals that content is current and maintained. Especially for B2B expert content in the DACH region, where professional authority counts, these details make expertise and accountability explicit instead of leaving them to chance interpretation.

Image and media optimization with structured data

Visual and multimodal search are gaining importance, and structured data covers them too. ImageObject marks up images with properties such as license, creator, and description. Meaningful alt texts provide the textual meaning of an image, which supports both accessibility and machine understanding. Caption markup links an image and its caption unambiguously. For multimodal AI answers that combine text and images, this markup provides context that a bare image does not carry. Truthfulness applies here as well: the markup has to match the image actually shown.

Measurement and monitoring

Classic click metrics alone no longer capture the value of schema and GEO. The click decline caused by AI Overviews in Germany and Austria, minus 17.8 percent with nearly stable impressions, shows that visibility and clicks are diverging. If you only measure clicks, you underestimate your actual reach.

Up-to-date monitoring covers several levels. AI citation tracking observes whether and how content appears in AI answers. Analyzing Google Search Console data after the AI Overviews rollout in Austria and Germany separates impressions from clicks and makes the effect visible. Impressions, brand mentions, and presence in the answer layer are gaining importance as KPIs beyond the click. Realistic expectations still matter: since, according to the Ahrefs analysis, schema does not produce a meaningful increase in citations, it belongs in monitoring as an eligibility foundation, not as the driver of a citation curve. The success of schema shows in correct display and machine interpretability, the success of GEO in content visibility.

Going further

Schema markup on-page is a foundation, not a miracle cure. It secures eligibility for enhanced displays, creates entity clarity, and makes content unambiguously interpretable for machines. The evidence shows it is not a direct lever for more AI citations. The bigger lever in the answer layer is substantive content in the GEO sense: documented claims, clear sources, high entity density. For B2B in the DACH region, the reliable strategy is therefore a solid JSON-LD base with the core types, clean semantic HTML underneath, citable content as the actual visibility driver, and monitoring that measures impressions and presence, not just clicks.