Structured data — machine-readable labels added to a page so a system can tell what the content is, instead of inferring it from prose.
What Google actually says it does
Two things, neither of which is ranking. It helps Google understand the content of a page, and it makes certain pages eligible for rich results in search.
Google is unusually direct about the AI question: “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add.” It still recommends it as part of ordinary SEO, for rich result eligibility.
So anyone selling you schema markup as an AI ranking factor is selling something the vendor itself has publicly denied. That does not make the work worthless — it makes the pitch dishonest.
Source: Google Search Central, Introduction to structured data and the AI optimization guide.
Where it earns its keep
Ambiguity is what stops a machine repeating a fact about you confidently. Structured data removes ambiguity from the facts a buyer actually needs: who you are, where you are, what you do, when you are open, what it costs, what people said.
The working test: if a fact matters to a buyer and only a human reading prose could extract it reliably, mark it up.
Related: entity resolution, knowledge graph, AI Overviews.