Schema.org has over 800 types and thousands of properties. Answer engines do not read them all with equal weight. Forty properties across eight types carry most of the signal. Ship those forty correctly and the rest of the Schema.org tree is noise that bloats payload without moving citations. The broader AEO framework this schema work fits inside lives in Answer Engine Optimization Explained.
This post is the tactical guide. It covers the eight types that matter, the forty properties inside them that move citation decisions, the cross-referencing shape a coherent graph takes, and the linter workflow that gates every ship. The target reader is an operator on a sub-DR-20 site who wants the shortest path between JSON-LD and a Perplexity citation with attribution.
Which Schema.org Types Matter Most for Answer Engines?
Person. Organization. BlogPosting. SoftwareApplication. FAQPage. HowTo. BreadcrumbList. WebSite. Every answer-engine visible surface for a sub-DR-20 brand reduces to a graph of these eight. Everything else, Event, Recipe, Course, Product, Review, only matters if the brand actually operates in those verticals. For a thought-leader brand plus a product brand, these eight are sufficient and complete. Those eight describe what a site is, and letting an agent act on it is a separate layer. What WebMCP is covers the experimental browser interface for that.
Which 40 Schema.org Properties Carry the Most Citation Weight?
Heavy-weight properties do the citation work. Person.sameAs. Person.knowsAbout. Organization.sameAs. Organization.founder. BlogPosting.author. BlogPosting.headline. BlogPosting.datePublished. SoftwareApplication.creator. SoftwareApplication.applicationCategory. WebSite.publisher. Medium-weight properties supply context an engine uses when disambiguating between candidate entities. Light-weight properties fill out the graph and show the brand cared enough to structure its data.
The rule is not more properties. The rule is the correct forty, cross-referenced. How much those forty matter depends on the engine, and Perplexity versus ChatGPT citation rules shows schema weighing more for ChatGPT than for Perplexity.
Why Does Graph Shape Beat Property Count for Schema.org?
A coherent graph is worth more than a long flat list. Person.creator points at SoftwareApplication’s @id. SoftwareApplication.publisher points at Organization’s @id. BlogPosting.author points at Person’s @id. Every @id resolves. Engines that walk the graph from any node can reach every other node in three hops. This is the structural signal of an entity that has thought about itself.
§4, The linter is the spec
Do not write JSON-LD by hand without running it through the Schema.org validator and Google Rich Results test. If both pass, engines will parse it. If either fails, some engines silently discard the whole block. The workflow is five steps. Draft. Lint. Cross-ref gate. Ship. Measure. The 90-day AI visibility roadmap shows where that ship step falls in a full program. It also covers what has to happen before that step.
Which Schema.org Types Should You Skip?
Skip WebPage. Skip SiteNavigationElement. Skip Article unless you have a strong reason not to use BlogPosting. Skip AggregateRating unless you actually aggregate ratings. Every type you add is a maintenance commitment. Broken schema is worse than missing schema because it shakes engine confidence in the parts that are correct.
§6, Measurement
The observable signal is citation volume and quote length in answer engines, plus rich-result eligibility in Google. If JSON-LD is doing its job, you will see citation count rise per named entity inside a quarter, not from a single post going viral, but from engines consistently resolving queries to the same canonical entity.
Bridge
citability.dev runs a schema-linter gate on any URL and reports which of the forty properties are missing, drifted, or pointing at a dangling @id. Ship through that gate and the forty are coherent by default. You can see a worked example in the live citability.dev entity map, which renders the entity-attribute pairs an AI crawler extracts from a coherent graph.
Forty properties across a real site is roughly a week of careful work. Before you spend it, run the free AEO audit on the domain. It scores crawler access, content extractability, AI metadata, and answer-ready formatting. It also reports structured data and llms.txt without scoring them. That tells you whether engines can reach and read your pages before you start on markup.
