Schema markup helps AI systems understand your content, but the Ahrefs study published in May 2026 found it doesn’t directly increase AI citations. This guide explains what schema actually does for AI Overview visibility, why 53% of AI-cited pages include structured data without that causing the citations, and where to focus GEO investment instead.
Schema markup helps AI systems understand your content, but the Ahrefs study published in May 2026 found it doesn’t directly increase AI citations. That finding surprised a lot of SEO teams who had been told schema was the unlock for AI Overview visibility. The evidence tells a more useful story.
Key takeaways
- Ahrefs tracked 1,885 pages adding JSON-LD schema and found no meaningful citation uplift across Google AI Overviews, AI Mode, or ChatGPT
- 53% of AI-cited pages already include structured data, making schema a floor condition rather than a citation driver
- Schema markup is necessary groundwork for entity recognition and AI understanding, even when it doesn’t directly produce citation gains
- Organisation schema and entity linking, not page-level schema alone, produce measurable improvements in AI Overview visibility
Schema is one of those topics where the industry consensus ran ahead of the evidence. The teams we work with at FirstMotion had often already implemented schema across their sites before coming to us, and still had near-zero AI citations for their most important queries. Our ContextualJourney™ platform maps this gap at the entity level, showing where AI systems lose confidence in a brand’s identity before they ever evaluate the content. Schema is part of the foundation, but it’s a long way from the whole story.
How schema markup helps AI search engines understand your content
Schema markup is a specific code vocabulary added to a website’s HTML, typically as a JSON-LD code snippet, that describes content to search engines and AI systems in machine-readable terms. JSON-LD is the recommended implementation format according to Google Search Central, and it’s what AI engines including OAI-SearchBot and PerplexityBot process at crawl time. Rather than leaving AI models to infer meaning from unstructured text, schema markup defines entities, relationships, and context explicitly.
In March 2025, both Google and Microsoft confirmed publicly that they use schema markup for their generative AI features. Krishna Madhaven from Microsoft described schema as a “steering” mechanism that builds AI confidence in the correct answer for a user’s query. Schema markup also supports voice assistants and semantic search by clarifying query nuances and removing ambiguity at the ingestion stage.
There are over 800 schema types available covering various types of content, from articles and businesses to products, events, and how-to written guides. The schema types that matter most for AI search are covered in the section below.
What the Ahrefs study found: schema markup and AI Overviews
Ahrefs published a controlled study in May 2026 tracking 1,885 pages that added JSON-LD schema between August 2025 and March 2026, matched against 4,000 control pages with similar AI citation histories. Citation changes were measured 30 days before and after schema addition across Google AI Overviews, Google AI Mode, and ChatGPT using a matched difference-in-differences methodology that strips out platform-wide trends.
| Platform | Citation change | Verdict |
|---|---|---|
| Google AI Mode | +2.4% | Statistically indistinguishable from noise |
| ChatGPT | +2.2% | Statistically indistinguishable from noise |
| Google AI Overviews | -4.6% | Small but statistically significant; not confidently attributable to schema |
Four separate statistical tests all pointed the same way. Adding schema markup produced no major uplift in citations on any AI platform.
The study’s scope constraint matters. Every page in the sample already had a meaningful AI Overview citation baseline before schema was added. The finding is that adding schema to a page already on the AI citation track doesn’t move the needle. It says nothing about how schema performs for pages with no citation baseline at all. In our entity audits, we consistently find brands in exactly that position. For those brands, schema is still the right first step.
Why 53% of AI cited pages use structured data
According to Ahrefs’ analysis of 6 million URLs, 53% of AI-cited pages include structured data, and pages with structured data are almost three times more likely to appear in AI Overviews than pages without it. Both facts are accurate, and neither contradicts the other. Well-maintained sites with high quality content, strong domain authority, and genuine topical expertise tend to implement schema. The correlation is a byproduct of those sites, not caused by the schema itself.
In practice, the brands we audit with strong AI citation rates almost always have schema in place alongside strong entity presence. The schema didn’t cause the citations, but its absence would have introduced friction the other signals couldn’t fully compensate for. This is the correlation-causation gap the Ahrefs controlled study was designed to test.
When you strip out other factors by matching treated pages against equivalent control pages, schema’s independent contribution disappears. For B2B software brands, schema is a baseline hygiene requirement rather than a citation lever. Implementing it removes unnecessary ambiguity for AI tools. Deploying it sitewide and expecting a step-change in generative results produces the same outcome the Ahrefs study found.
Schema types for AI search: Article, Person schema and FAQ schema
Not all schema types carry equal weight for AI search. The types that matter most are the ones that build entity clarity and content extractability for AI users, not the ones that produce rich results in traditional search.
| Schema type | What it communicates to AI | Why it matters for generative results |
|---|---|---|
| Organisation | Brand identity, category, service areas, sameAs URLs | Resolves entity disambiguation across AI platforms |
| Person | Author credibility, affiliations, published work | Establishes author bio signals AI models evaluate for E-E-A-T |
| Article | Content type, publication date, authorship | Produces accurate AI generated answers by giving AI models structured metadata |
| FAQ | Question and answer pairs in extractable format | Improves content extractability for AI Overviews even without FAQ rich results |
| Product / SoftwareApplication | Features, pricing, availability | Describes product capabilities accurately in AI generated answers |
Google restricted FAQ rich results to authoritative government and health websites in August 2023, with full deprecation completed in May 2026. FAQ schema still improves content extractability for AI systems. Keeping schema markup updated to match visible page content is increasingly important: AI models compare structured data against rendered HTML, and mismatches reduce citation confidence rather than building it.
Organisation schema and entity recognition for AI systems
Organisation schema is the single most strategically important schema type for B2B software brands focused on AI search visibility. It connects all digital signals associated with a business into a single, unambiguous entity that AI systems can identify, verify, and trust as a source. The sameAs property is where the real work happens: it links your website entity to your Wikipedia page, Wikidata entry, LinkedIn profile, and other authoritative URLs that AI platforms use as reference points for the same entity.
In our audits, incomplete or missing sameAs properties in Organisation schema are one of the most common fixable entity gaps we find, and one of the fastest to resolve. Schema App’s entity linking study showed a 19.72% increase in AI Overview visibility after implementing entity linking that connected on-page entities to authoritative external knowledge bases including Wikipedia, Wikidata, and Google’s Knowledge Graph. That result came from connected schema with entity linking across authoritative references, rather than from adding basic JSON-LD schema types to existing pages.
Linked data, knowledge graph and AI visibility
Most of the apparent contradiction in the research comes down to one distinction: schema markup versus connected schema with entity linking. Adding JSON-LD to a page tells AI systems what that page is about. Connecting page entities to external reference databases via sameAs references tells AI systems that the entity on this page is the same entity they already know from Wikipedia and Wikidata. AI models can then resolve the entity to a known identity rather than treating it as an ambiguous text string.
Google’s Knowledge Graph acts as the reference layer AI platforms draw from when forming their understanding of entities. A brand with a verified Knowledge Graph entry linked to its schema markup enters generative results with significantly higher confidence than a brand relying solely on page content. Building this connection takes longer than deploying JSON-LD. It’s also what the evidence shows actually moves AI Overview visibility.
Schema markup strategy for AI search in 2026
A practical schema strategy treats markup as entity infrastructure rather than a citation shortcut. Four priorities in sequence:
- Deploy Organisation schema sitewide with complete sameAs references to Wikipedia, Wikidata, and LinkedIn. This produces the strongest entity recognition gains across all major AI platforms
- Implement Article schema on all published content with accurate authorship, publication dates, and category. Keep it updated to match visible page content; mismatches reduce AI confidence
- Add Person schema for named authors with sameAs references to LinkedIn profiles and published work. Author bio signals are increasingly important to AI models evaluating source credibility
- Use FAQ schema on question and answer content even without FAQ rich results. Run all schema through Google’s Rich Results Test to confirm accuracy before publishing
For B2B software brands, SoftwareApplication schema is also worth implementing for product pages. It gives AI models the structured product data they need to describe your product accurately in generative answers.
Rich results and what schema still delivers for AI search
Schema markup’s direct value for traditional search results remains real. Rich snippets including star ratings, pricing, and review counts still appear for correctly implemented schema and still produce higher click-through rates than standard links. For B2B software brands, this traditional search value alone justifies schema investment.
Earned authority, consistent entity presence, and content that answers users’ queries at the passage level drive AI citations. Schema supports all three but doesn’t replace any of them. Redirect GEO investment beyond schema into earned media and entity signals, the factors the evidence shows actually determine whether AI platforms cite your brand.