Schema Markup for AI Optimization Did Not Improve AI Citations

For months, the SEO community has been operating on a widely held assumption: add structured data, get more visibility in AI-generated answers. It sounds logical. AI systems need to understand content. Schema markup explains content. Therefore, schema helps AI. Clean narrative. Wrong conclusion – at least according to a controlled study from Ahrefs.

I’ve been watching this assumption spread through the industry unchecked, repeated in conference talks, agency audits, and client deliverables as though it were proven fact. It isn’t. And now there’s rigorous data to back that up. Let me walk through what the study found, what it missed, and what this means for your actual SEO and AI visibility strategy.

How Does Schema Impact AI citations? What the Ahrefs Study Found

When Ahrefs ran a controlled experiment – adding schema to pages that never had schema markups and comparing them against matched control pages – schema addition produced no measurable positive impact on AI citation frequency over a 30-day observation window.

The methodology matters here. Ahrefs didn’t just look at correlations; they ran a difference-in-differences analysis using their Brand Radar tool and Agent A. They matched treated pages (those receiving schema additions) against control pages without JSON-LD, then tracked citation changes across four separate tests. The result: no statistically significant improvement from adding schema markup.

What they did observe was a 4.6% decline in AI Overview citations – a number flagged as statistically notable. Both the treated and control groups were declining before schema was introduced, but pages that received schema declined slightly faster, losing approximately 12 additional daily citations compared to controls. Ahrefs was careful not to declare causation here, noting the decline could be coincidental.

“Correlation between schema and AI citations exists – but it’s almost certainly explained by site quality, not structured data itself.”

The Correlation vs. Causation Problem the Industry Keeps Getting Wrong

Here’s the thing that frustrated me most about how this topic has been discussed: the original correlation finding – that AI-cited pages are 3x more likely to have JSON-LD – was never evidence that schema causes AI citations. It was always evidence that well-maintained, high-quality websites tend to implement schema as part of broader technical scope.

Sites that use JSON-LD correctly also tend to:

• Have faster page speeds and cleaner technical infrastructure
• Publish more structured, authoritative content
• Earn more backlinks and demonstrate stronger E-E-A-T signals
• Get crawled and indexed more efficiently
• Maintain consistent content updates and semantic topical coverage

The schema is a marker of a well-run site – not the mechanism driving AI citations. Conflating the two is one of the more persistent mistakes I see in this industry right now.

What Schema Markup for AI Search Actually Does (and Doesn’t Do)

Schema markup for AI search helps AI systems interpret structured information – business hours, product details, FAQ content, author credentials – but the Ahrefs study suggests it does not independently increase the probability of being cited in AI Overviews for pages already recognized by AI systems. Its primary documented benefits remain in rich results, knowledge graph reinforcement, and crawlability.

What Schema Markup Genuinely Helps With

• Rich Results Eligibility: Product schema, and Review schema can trigger enhanced SERP features that improve click-through rates.
• Knowledge Graph Reinforcement: Entity-level schema (Organization, Person, LocalBusiness) helps Google’s Knowledge Graph understand who and what you are, which has downstream effects on brand understanding.
• Crawling Efficiency: For large or complex sites, structured data can help crawlers navigate content relationships more effectively.
• Semantic Clarity: Schema provides explicit signal about content type – recipe, article, event, product – reducing ambiguity in content interpretation.

What Schema Markup Does Not Do

• It does not directly increase AI Overview citation frequency for already-recognized pages.
• It does not substitute for content depth, authority, or topical relevance.
• It does not guarantee visibility in ChatGPT, Gemini, Perplexity, or Claude responses.
• It does not override weak E-E-A-T signals or thin content.

The Real Mechanism Behind AI Citations

I want to be direct about something: the AI citation problem is fundamentally a content authority problem, not a technical markup problem. When Google’s AI Overviews, Perplexity, or ChatGPT pull from your page, they’re doing so because your content satisfies specific criteria that have very little to do with JSON-LD.

Based on everything I’ve observed in this space, AI systems prioritize content that:

1. Directly answers specific questions – AI systems are trained to extract precise answers, not pages that dance around a topic.
2. Demonstrates genuine expertise – Depth of explanation, use of industry-specific terminology, and author credentials matter significantly.
3. Is highly linked and referenced – Backlink authority remains a proxy for trustworthiness across both traditional and AI search.
4. Is consistently accessible and fast – Technical performance affects crawl frequency and freshness signals.
5. Contains unique information or perspective – Generic content that mirrors a dozen other pages provides no incremental value to AI systems assembling composite answers.

“The pages getting cited most frequently in AI Overviews aren’t winning because of their markup. They’re winning because they’ve built genuine topical authority and answer questions better than anyone else.”

Critical Gaps in the Ahrefs Study Worth Understanding

I respect the rigor of what Ahrefs did here, but it’s important to read their own acknowledged limitations carefully rather than walking away with oversimplified conclusions.

The Study Only Covered Already-Cited Pages

Every page in the dataset had 100 or more AI Overview citations before the schema was added. This is a significant constraint. The study tells us schema doesn’t meaningfully increase citations for pages already visible to AI – it says nothing about whether schema might help pages that aren’t yet being cited get discovered and included. That’s an entirely different research question, and one that remains genuinely open, even though, from the study, we can probably conclude that schema isn’t doing much when it comes to AI citations, since pages without schema markup were already being cited.

All Schema Types Were Pooled

FAQ schema, Article schema, Organization schema, and Product schema all behave differently and serve different purposes. Grouping them together in analysis could mask meaningful variation. A study that isolates the effect of, say, FAQ schema on question-based AI queries would be far more informative.

The 30-Day Window May Be Too Short

Structural changes to how AI systems index and weight content signals may operate on longer timescales than 30 days. A study observing 90 to 180-day windows after schema implementation might reveal patterns invisible in shorter observation periods.

Simultaneous Page Changes Introduce Noise

Real-world SEO is messy. When schema is added to a page, other changes often happen concurrently – content edits, internal link updates, image optimizations. Isolating the pure effect of schema addition in a live environment is genuinely difficult, and Ahrefs acknowledged this limitation honestly.

Myths vs. Facts: Schema Markup for AI Optimization

Where Schema Markup Still Belongs in Your Strategy

None of this means you should abandon structured data. That would be the wrong takeaway. Schema markup remains a foundational technical SEO practice – the Ahrefs findings simply clarify its role rather than eliminate it.

My practical position: implement schema because it’s technically sound, supports rich results, reinforces entity understanding, and reflects the kind of site quality that correlates with long-term AI visibility. Just don’t implement it as a shortcut to AI citations, because that’s not what it does.

Schema Types Still Worth Implementing

• Organization / LocalBusiness: Foundational for entity recognition and knowledge graph inclusion.
• Article / BlogPosting: Helps establish authorship and content freshness signals.
• Product / Offer: Essential for e-commerce visibility in both traditional and AI-assisted shopping queries.
• Person: Critical for author E-E-A-T signals, especially for YMYL content categories.
• BreadcrumbList: Helps establish content hierarchy for large sites.

What Actually Moves AI Citations: A Framework

After observing how AI search systems behave across hundreds of sites and query types, I’ve developed what I think of as the AI Citation Authority Framework. It’s not about any single technical lever – it’s about compounding signals.

The AI Citation Authority Framework

1. Topical Concentration: Own a topic cluster deeply rather than covering everything superficially. AI systems pattern-match to established authorities within specific domains.
2. Answer Precision: Structure content to deliver direct, quotable answers within the first 100 words of a section. Vague, hedging language rarely gets extracted.
3. Demonstrated Experience: First-person observations, specific data points, and unique perspectives differentiate your content from the homogenized middle of the web.
4. Citation Equity: Being referenced by other high-authority pages sends trust signals that influence AI system training data weighting.
5. Entity Clarity: Make it unambiguous who wrote the content, what organization it represents, and what the page is definitively about – through content signals, not just markup.
6. Technical Accessibility: Fast load times, stable URLs, clean crawl paths, and proper canonicalization ensure AI crawlers can access and re-access your content consistently.

A Practical Observation From Working in This Space

What I’ve noticed consistently is that clients who obsess over schema implementation while neglecting content depth tend to see minimal AI visibility gains. Meanwhile, sites that publish genuinely useful, well-structured content written by actual experts – even with minimal schema – frequently appear in AI-generated answers across Perplexity, Google AI Overviews, and ChatGPT’s browsing mode.

The Ahrefs study validates this observation with actual data. Structured data is table stakes for a technically healthy site. It is not the mechanism for AI search optimization. The industry needs to stop conflating the two.

“If you’re relying on schema to compensate for weak content, you’re optimizing the container while ignoring what’s inside it.”

What This Means for SEO Strategy Going Forward

The Ahrefs findings suggest that schema markup for AI optimization should be treated as a supportive technical practice rather than a primary AI visibility lever. Strategies focused on content depth, genuine topical authority, E-E-A-T signals, and earning citations from other credible sources are likely to produce more measurable AI citation results than structured data additions alone.

Recommended Priorities

• Audit your content for answer precision – can AI systems extract a clear, direct answer from your pages?
• Build topical depth across your core subject areas rather than chasing individual keywords.
• Strengthen author entity signals – bylines, author pages, consistent publishing history, and external credentials matter.
• Earn coverage and links from established publications in your industry; these citation signals translate into AI trust signals.
• Implement schema correctly as part of technical hygiene, not as an AI optimization shortcut.
• Monitor AI Overview and AI citation visibility separately from traditional ranking metrics – they respond to different signals.

Frequently Asked Questions

Does schema markup improve AI Overview citations?

Based on Ahrefs’ controlled study adding JSON-LD schema to pages already receiving AI Overview citations produced no measurable improvement in citation frequency over a 30-day period. The correlation between schema and AI citations that exists in aggregate data appears to reflect overall site quality rather than a direct causal relationship between structured data and AI visibility.

Why do AI-cited pages have more schema markup if schema doesn’t help?

Pages frequently cited in AI Overviews tend to be high-quality, well-maintained sites that implement schema as part of broader technical excellence. The schema is a co-occurring characteristic of these sites, not a driver of their AI visibility. Well-resourced sites do more things right simultaneously – schema is one of them, but it’s not the variable causing AI citations to increase.

Should I still implement schema markup for AI search optimization?

Yes, but with accurate expectations. Schema markup remains valuable for rich results eligibility, knowledge graph entity recognition, and technical crawlability. It should be implemented as part of sound technical SEO practice. However, it should not be positioned as a primary strategy for increasing AI citations or AI Overview appearances – the evidence doesn’t support that claim.

What does actually improve AI citation frequency?

The strongest drivers of AI citation frequency appear to be content authority signals: topical depth and concentration, direct and precise answer structures, strong E-E-A-T indicators (demonstrated experience, authorship credentials, institutional trust), backlink authority from credible sources, and consistent technical accessibility for AI crawlers. These content and authority factors outweigh any observed effect from structured data additions.

Does schema markup help pages that aren’t yet cited by AI?

The Ahrefs study did not test this scenario – their dataset was limited to pages already receiving 100 or more AI Overview citations. Whether schema helps previously uncited pages get discovered and included in AI-generated answers remains an open research question. It’s plausible that schema could assist with initial crawling and indexing for lower-visibility pages, but this specific question requires further study with a different experimental design.

Final Thoughts

The Ahrefs study is one of the more methodologically serious pieces of research the SEO industry has produced on AI search visibility, and it deserves to be taken seriously rather than dismissed or cherry-picked. Its core finding – that schema markup for AI does not move the needle on citations for already-recognized pages – challenges a narrative that has been circulating without adequate scrutiny.

That doesn’t mean structured data is worthless. It means we need to be honest about what it actually does versus what we’ve been assuming it does. Schema supports rich results, entity recognition, and technical clarity. Content authority, answer precision, and earned trust drive AI citations. These are different levers, and treating them as equivalent has led a lot of practitioners to invest time and budget in the wrong places.

My recommendation is simple: keep implementing schema correctly because it’s the right technical practice, but build your AI visibility strategy on the foundations that demonstrably matter – depth, authority, precision, and trust.

If you want a clear-eyed assessment of what’s actually driving or limiting your site’s visibility in AI-generated search results, I’m happy to dig into the specifics. No assumptions, no shortcuts – just an honest look at the signals that matter.