TL;DR

Ahrefs tracked 1,885 pages that added JSON-LD schema markup between August 2025 and March 2026, compared them against ~4,000 control pages, and measured AI citation changes across three platforms. Google AI Overviews citations dropped 4.6%. AI Mode went up 2.4%. ChatGPT went up 2.2%. None of those numbers are statistically significant. Schema markup, the single most recommended GEO tactic in 2025-2026, doesn't move AI citations for pages that are already being cited.

Key Takeaways

Schema didn't help on any platform. Across Google AI Overviews, AI Mode, and ChatGPT, adding JSON-LD produced no meaningful citation uplift.
AI engines read visible HTML, not hidden markup. A complementary searchVIU experiment found that ChatGPT, Claude, Perplexity, Gemini, and Google AI Mode extracted only rendered page content. JSON-LD, hidden Microdata, and hidden RDFa were completely ignored.
The study has a real limitation. Every page in the dataset already had 100+ AI Overview citations before schema was added. For brand-new sites with zero AI visibility, schema may still help with crawling and parsing. But for established brands trying to increase citations: it's not the lever.
Five signals actually move AI citations in 2026. Brand authority, third-party mentions, content freshness, entity clarity, and citation density from independent sources.

What the Study Actually Found

The Ahrefs blog post, published May 13, 2026, lays out the numbers plainly. Researchers selected pages that added JSON-LD schema between August 2025 and March 2026. Each page had at least 100 AI Overview citations in February 2025, giving a solid pre-treatment baseline. The control group (~4,000 pages) had similar citation volumes but no schema changes.

Results across three platforms:

Google AI Overviews: -4.6% (a slight decline vs. control)
Google AI Mode: +2.4% (within random variance)
ChatGPT: +2.2% (within random variance)

The searchVIU experiment adds context. When researchers tested direct content retrieval by five AI engines, every single one extracted visible HTML only. JSON-LD sat untouched in the page head. The structured data that GEO guides have been pushing as essential? AI models don't read it.

Why This Matters: Schema Is GEO Advice Number One

Pick any GEO or AEO guide published in the past 18 months. Similarweb's AI search playbook, Connor Kimball's citation optimization framework, Yotpo's brand visibility checklist, ROI Revolution's GEO primer, mekaa's AEO toolkit: all of them list schema markup in their top five recommendations. Some put it at number one.

That advice was always logical. Structured data helps machines parse content. If AI engines are machines parsing content, schema should help them cite you more often. Clean reasoning, zero evidence. Ahrefs just provided the evidence, and it points the other direction.

This isn't unique. We saw the same pattern with llms.txt: the industry pushed a tactic hard, adoption stayed low, and measurable impact remained unclear. Two pillars of the 2025-2026 GEO playbook, both lacking data support.

The Fair Caveat: When Schema Probably Still Helps

I Love SEO published a counter-take worth reading. Their argument: Ahrefs tested pages that were already heavily cited. For those pages, adding schema is optimization on top of optimization. The real question is whether schema helps invisible pages get their first citations.

That's fair. Schema still serves three purposes that aren't about AI citation uplift:

Crawl efficiency. Structured data can speed up how search engines parse your content type, author info, and publication date.
Rich results in traditional search. FAQ schema, HowTo schema, and review schema still trigger rich snippets in Google's organic results.
Entity disambiguation. For brands with common names, Organization and Product schema helps search engines (including AI) understand which entity you are.

The takeaway: if your site has zero schema, adding it is still reasonable hygiene. But if you're already being cited by AI engines and want to increase your citation share, schema isn't where to spend your time.

What Actually Moves AI Citations in 2026

Five levers have data behind them:

1. Brand authority from independent sources. Microsoft's published grounding signals for Copilot and Bing AI don't mention schema once. They weight authoritative third-party mentions, brand recognition patterns, and entity consistency across the web. Get cited by others before trying to cite yourself.

2. Third-party mentions and earned media. Pages that accumulate mentions on Reddit, industry publications, and review sites see higher AI citation rates. This is the earned media GEO playbook: your brand's citability depends on what others say about you, not what you say about yourself.

3. Content freshness. AI engines heavily favor recently updated content. A fresh-updates badge on your key pages, with genuine new data rather than cosmetic date bumps, moves citation rates measurably.

4. Entity clarity and pre-query signals. AI visibility gets shaped before the user types a query. Your brand's entity graph, the connections between your brand name, product category, key people, and industry terms, determines whether AI engines consider you a candidate for citation at all.

5. Citation density across platforms. Being cited once in one AI engine isn't enough. Citation share across platforms, from Google AI Overviews to ChatGPT to Perplexity, compounds. Each citation reinforces your entity in training data and retrieval indices.

Every one of these levers is measurable. Schema markup uplift wasn't, because there was nothing to measure.

Measure Before You Optimize

The Ahrefs study is a reminder: GEO tactics need evidence, not intuition. Before adding schema, before rewriting content, before any optimization, the first step is knowing where your brand actually stands in AI citations.

Which queries cite you? Which cite your competitors instead? Where is your citation gap widest?

Be Recommended tracks exactly that. Audit your brand's AI citation share across engines, identify the signals that actually move it, and stop investing in tactics that the data says don't work.

FAQ

Does schema markup help with AI search visibility?

According to Ahrefs' May 2026 study of 1,885 pages, adding JSON-LD schema produced no meaningful uplift in AI citations on Google AI Overviews, Google AI Mode, or ChatGPT. Schema still helps with traditional rich results and crawl efficiency, but it doesn't increase how often AI engines cite your content.

What did the Ahrefs schema study test?

Ahrefs compared 1,885 pages that added JSON-LD schema (August 2025 to March 2026) against approximately 4,000 control pages. They measured citation rate changes across three AI platforms. Results ranged from -4.6% to +2.4%, none statistically significant.

What actually improves AI citations in 2026?

Five evidence-backed signals matter more than schema: brand authority from independent sources, third-party mentions and earned media, content freshness with genuine updates, entity clarity in your brand's knowledge graph, and citation density across multiple AI platforms.

Should I remove schema from my site?

No. Schema still serves purposes beyond AI citations: rich results in Google organic search, crawl efficiency, and entity disambiguation. Keep it as part of your technical SEO baseline, but don't treat it as an AI visibility lever.

Ahrefs Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved.