Schema Markup Added to 1,885 Pages Produced No AI Citation Gains, Study Finds

The Situation

That schema markup line item on your web developer's invoice probably didn't buy you any new AI citations. Ahrefs tracked 1,885 pages that added JSON-LD schema markup between August 2025 and March 2026, and citations from Google AI Overviews, AI Mode and ChatGPT barely moved. That undercuts one of the most repeated pieces of SEO advice in the AI-search era: that structured data is a citation lever. For a dealer paying a developer to bolt structured data onto the website as an AI-visibility play, the study puts that spend in the housekeeping column.

The Facts

Ahrefs, an SEO analytics firm, built its dataset by pulling pages cited in Google AI Overviews through its Brand Radar tool, then checking crawl history for the date each page's HTML picked up a JSON-LD schema tag. That produced 1,885 treated pages. Ahrefs compared the treated pages with 4,000 control pages from other domains that had never added JSON-LD and were being cited at similar levels beforehand. It then measured citation counts 30 days before and after the schema went live.

The correlation SEOs have leaned on for years is real: Ahrefs' original scan of 6 million URLs found AI-cited pages were almost three times more likely to carry JSON-LD than uncited ones. The follow-up test found that gap doesn't hold up as cause and effect. No platform showed a meaningful citation increase after schema was added. AI Mode and ChatGPT treated pages ran only slightly ahead of controls, a gap Ahrefs says is small enough to be random noise across thousands of URLs. AI Overviews moved the other way: citations on treated pages fell 4.6% relative to controls, a result Ahrefs calls statistically significant (the odds of a gap that size happening by chance are about 1 in 2,500). Four separate tests, including a difference-in-differences model and a week-by-week event study, pointed to the same conclusion.

One limitation matters most for a dealer weighing whether to bother. Ahrefs found every page in the study already had 100+ AI Overview citations in February 2025, before any schema went on, meaning the pages tested were already inside the AI consideration set. Whether schema helps a page get crawled and indexed in the first place is not yet confirmed; Ahrefs says its data can't speak to that question.

A separate searchVIU experiment cited in the study found none of five AI systems tested pulled schema data when fetching pages live. They read only the visible HTML, though that doesn't rule out schema mattering earlier, during training or indexing.

Business Impact

Our read of Ahrefs' data: if schema is not the lever, the page content is. For a propane dealer that means specific text about service area, delivery windows, safety steps and account questions, the kind of detail an AI system can lift straight into an answer. Ahrefs did not test propane sites. Schema still feeds Google's classic rich results and knowledge panels. This study found it isn't the AI-visibility shortcut some agencies have pitched. What this means for your business: 1. Budget for content updates on service-area, delivery-window and account-management pages before a schema-only project if AI citations are the goal. 2. Don't expect an AI-visibility bump from a schema update alone; measure it against rich-results eligibility instead. 3. If a page has never been cited by AI, schema may still help it get crawled and indexed, a different problem than winning citations for pages AI already sees. 4. Push any vendor selling schema as an 'AI SEO fix' to show platform-level evidence, not just the correlation.

Key Takeaways

• Treat schema as housekeeping for Google's rich results, not as an AI-visibility project.
• Put the next web dollar into clear, specific answers on the service-area, delivery and safety pages.
• Ask any vendor pitching schema as an 'AI SEO fix' for platform-level evidence, not the old correlation.
• Ahrefs studied pages AI already cited heavily; for a page AI has never cited, schema's role in crawling and indexing is still an open question.

Action Steps

1. 1.Pull your site's service-area, delivery-window and account-management pages this week and check whether they answer questions in plain, specific text an AI system could quote directly.
2. 2.Call your web developer or SEO vendor this week and ask them to show platform-level evidence, not just the schema-versus-no-schema correlation, before you pay for a schema-only AI-visibility project.
3. 3.Check Google Search Console for rich-results eligibility on key pages before your next site update, since that's what JSON-LD schema still reliably supports.
4. 4.If a page has never been cited by AI Overviews, AI Mode or ChatGPT, keep its schema in place; Ahrefs' data doesn't rule out schema helping that page get crawled and indexed in the first place.

Competitive Advantage

Dealers who read past the schema hype and put their next web-development dollar into clearer service-area, delivery and safety content give AI systems the specific, quotable text they actually pull into answers. That's a cheaper, more durable edge than chasing a structured-data tactic Ahrefs' own data says doesn't move AI citations for pages already in the mix.