Lily Ray backs Pedro Dias, says content quality beats schema for AI search

Publishers chasing AI visibility need a reality check: heavy schema markup is not a shortcut into answer engines, and it is not a guarantee that LLMs will parse, connect, or cite your pages. Lily Ray reinforced that view after Pedro Dias took aim at a Semrush Technical GEO graphic that said schema and structured architecture “ensures” AI engines can parse and connect content.

Dias’ point cut straight through the current vendor pitch. Large language models work from tokens and patterns in text, not from a magical microdata layer that locks in inclusion. That is why the practical priority remains the same as it has been for years: write clear, information-rich copy, keep the site crawlable, and build an architecture that lets search systems understand what lives where. Schema still matters as a supporting signal, but it does not override weak content or promise a place in AI answers.

Ray has been saying versions of that for months. Her MozCon 2025 presentation, GEO, AEO, LLMO: Separating Fact from Fiction & How to Win in AI Search, framed those buzzwords as evolutions of longstanding SEO principles rather than a new discipline that replaces them. In a January 22, 2026 Moz AMA, she went further, saying publishers that depend on pageviews and affiliate traffic may struggle because AI Mode and AI Overviews are designed to keep users engaged rather than send them elsewhere. She also said AI-generated content may be technically allowed by Google, but quality, human expertise, and editorial oversight still matter in practice.

The traffic data gives that warning teeth. Pew Research Center reported on July 22, 2025 that Google users were less likely to click result links when an AI summary appeared. Google Search Central’s AI features guidance also points publishers back toward technical requirements and standard SEO best practices, saying AI features can help users find a website but stopping well short of promising special AI-only treatment or guaranteed inclusion.

That is the fault line in AI search right now. Google and some vendors still agree that structured data can help search systems understand content, but the debate is over the size of that role. Ray and Dias are pushing back against the idea that schema can buy certainty in a probabilistic system. For publishers, the safer bet is still the old one: make the page worth reading first, then use schema to support it.