- Ahrefs analyzed 1,885 pages with JSON-LD schema vs. control pages without schema
- Result: +2.4% citations in Google AI Mode, +2.2% in ChatGPT — both statistically insignificant
- In Google AI Overviews: -4.6% (a decrease, not an increase)
- Conclusion from Ryan Law, Head of Content at Ahrefs: "Probably not"
- Google removed FAQ rich results in 2026 — the context changes everything
Adding schema markup to boost AI citations? An Ahrefs study published May 16, 2026 shows this practice has virtually no effect. In AI Overviews specifically, it may even be slightly counterproductive.
The Ahrefs research team analyzed 1,885 web pages that had added JSON-LD schema markup, comparing them against control pages without schema. The result challenges a widespread belief in the GEO industry.
Is your GEO strategy built around schema markup? Get a free AI visibility audit — €250 to €1,800/month, agency-quality work at software-grade productivity.
The Study's Numbers
Here are the citation changes measured after adding schema markup to 1,885 pages (Source: Ahrefs, May 2026):
Statistically insignificant
Statistically insignificant
Measured decrease
Ryan Law, Head of Content at Ahrefs, is direct in his conclusions: "Does adding schema markup help your pages get cited in AI search? Probably not."
"Anything that can be spammed in SEO, will be spammed."
— Lily Ray, VP of SEO"The GEO industry is replaying early SEO, just faster."
— Joost de Valk, Founder of YoastContext: Google Removes FAQ Rich Results
This study arrives in an already disrupted landscape. Google definitively removed FAQ rich results from search results pages in 2026, restricting their display to recognized government and medical sites only (a restriction that began in 2023).
FAQPage schema, long considered a priority lever for SEO and GEO optimization, has now lost its value for traditional rich results. And the Ahrefs data suggests it doesn't offer a clear advantage for AI citations either.
The question becomes: if schema markup boosts neither rich results (removed) nor AI citations (negligible effect) — what is its value anymore?
What This Changes for Your GEO Strategy
The honest answer is that schema markup remains useful — but for different reasons than previously assumed.
What Schema Does NOT Do
- It does not guarantee AI Overview citations
- It does not replace genuine content expertise
- It is not a shortcut to AI visibility
What Schema Still Does
- Facilitates semantic understanding — search engines parse your content more easily
- Supports still-active rich results — Product, Recipe, HowTo, Event, Review remain
- Quality consistency signal — a technically structured site is still a quality indicator
- Accessibility for AI agents — Google-Agent and agentic crawlers still rely on structured data
The mistake to avoid: spending the next three months tagging pages in JSON-LD hoping to double your AI citations. The Ahrefs study shows that's not where your GEO strategy in 2026 should be focused.
What Actually Determines AI Citations
Cyrus Shepard's analysis on AI citation ranking factors (published May 2026) identifies more decisive signals than schema markup:
| Signal | Correlation with AI Citations |
|---|---|
| URL accessibility (no blocking) | Strong |
| Search ranking position | Strong |
| Fan-out rank (topical authority) | Strong |
| Intent-format match | Strong |
| JSON-LD schema markup | Not significant |
In other words: ranking well on Google remains the strongest predictor of being cited by AI systems. Classic SEO — content quality, topical authority, structure — stays the foundation of visibility in ChatGPT and AI Overviews.
Our Take: The GEO Industry Is Replaying SEO History
Joost de Valk is right. In SEO's early years, the industry chased technical shortcuts — keyword stuffing, meta keyword manipulation, link buying — before Google adjusted and brought content quality back to center stage.
History is repeating itself with GEO. Everyone's looking for shortcuts (schema, llms.txt, hidden AI prompts) when Google's own official GEO guide says "the only thing that matters is unique, valuable content."
The Ahrefs data confirms what serious practitioners already sensed: there is no technical shortcut to AI citation. What works is exactly what worked in SEO — genuine expertise, original content, topical authority.
What This Article Doesn't Cover
- The Ahrefs study covers generic schema markup (JSON-LD) — it didn't test all schema types (Product, Recipe, etc.)
- Results measure correlation with citations; impact on semantic comprehension by search engines is a separate question
- LLM algorithms evolve — these results reflect the state of May 2026
- The Cyrus Shepard citation factor study is preliminary and not peer-reviewed
Frequently Asked Questions
Should I remove schema markup from my site?
No. The Ahrefs study shows schema markup doesn't significantly improve AI citations — but it doesn't show it actively hurts them either (except for the -4.6% in AI Overviews, which warrants monitoring). Schema remains useful for still-active rich results (Product, HowTo, Recipe, Event) and for semantic comprehension. Recommendation: don't make it a priority for AI visibility, but don't remove it if it's already in place.
What replaces schema for boosting AI citations?
Based on data available in May 2026: (1) ranking well on Google — the strongest signal by far; (2) perfect technical accessibility (no blocking, robots.txt open to the right crawlers); (3) precise intent-format match; (4) topical authority development (fan-out rank). In short, classic SEO remains the foundation of AI visibility.
Did Google really remove FAQ rich results?
Yes. Google progressively reduced FAQ rich results since 2023 (limiting them to government and health sites), then fully removed them from SERPs in 2026. FAQPage schema remains technically valid, but it no longer generates enhanced displays in Google results for non-governmental sites. The residual value is semantic structuring for AI, but the Ahrefs study tempers expectations there as well.
Does the Ahrefs study apply to all AI platforms (ChatGPT, Perplexity, Claude)?
The Ahrefs study measured three platforms: Google AI Mode (+2.4%), ChatGPT (+2.2%), and Google AI Overviews (-4.6%). Perplexity and Claude are outside this study's scope. Similar results are likely since these AIs use comparable retrieval mechanisms (search-augmented generation), but that remains to be confirmed by platform-specific studies.
Growth and SEO content strategist, I founded Cicéro to help businesses build lasting organic visibility — on Google and in AI-generated answers alike. Every piece of content we produce is designed to convert, not just to exist.
LinkedIn