Schema Markup in the Age of AI Overviews

The Structured Data Question

Structured data and JSON-LD were created to help search engines understand content. But as AI Overviews replace traditional search results, many brands are questioning whether schema markup still matters. Our experiments provide a clear answer: it does—but not in the way most people think.

“Across 800 tested URLs, pages with correctly implemented JSON-LD were 3.1x more likely to be surfaced in AI Overview answer boxes than pages with identical content but no schema.”

Which Schema Types Still Matter

Not all schema is created equal in the eyes of LLMs. We found that certain types have a disproportionate impact on AI visibility:

• Article / NewsArticle — Signals freshness, authorship, and publication date—variables AI systems use to assess credibility.
• FAQPage — Perfectly suited for AI extraction. Questions and answers are often lifted verbatim into AI responses.
• HowTo — Strong performance in procedural queries. Each step is treated as a discrete, citable unit.
• Organization + Person — Critical for entity establishment. Tells AI systems who you are outside of any single page.

Common Implementation Mistakes

Our audit revealed several recurring errors that can actually hurt AI recognition:

• Marking up content that isn't visible on the page (Schema Spam).
• Using outdated Microdata instead of the preferred JSON-LD format.
• Missing critical properties like dateModified, which AI uses to gauge freshness.

A New Directive: Schema as AI Communication

Stop thinking of schema as a “search trick” and start thinking of it as a structured communication layer with AI systems. The more precisely you describe your content and entity in machine-readable terms, the more reliably AI systems can surface you in the right context.

Schema markup is not dead. It has simply evolved from a SERP feature hack into a fundamental layer of AI-era content infrastructure.