Semantic Triples for SEO: How to Write Facts Machines Can Read
Quick answer
A semantic triple is one fact written as subject, predicate, and object, such as "Example Bakery offers wedding cakes". Search engines and AI systems store knowledge as triples, so sentences that state facts in that order, and structured data that declares them, are the easiest parts of a page for machines to extract and reuse.
Key takeaways
- The W3C defines a triple as a subject, a predicate, and an object, conventionally written in that order.
- Knowledge graphs are large collections of triples; Google's launched in 2012 with more than 3.5 billion facts about and relationships between its objects.
- Sentences that state one fact in subject, verb, object order are the easiest for machines to extract.
- Structured data states triples explicitly: every property in a JSON-LD block is one.
- Passive voice, pronoun chains, and hedging hide the subject or the fact, which is what extraction needs.
What is a semantic triple?
A semantic triple is the smallest complete statement of a fact: a subject, a predicate that relates it to something, and an object. "Example Bakery", "offers", "wedding cakes". The idea comes from the web's data standards. The W3C's RDF specification defines a triple as "the subject, which is an IRI or a blank node; the predicate, which is an IRI; the object, which is an IRI, a literal or a blank node," and notes that a triple "is conventionally written in the order subject, predicate, object."
You do not need the technical terms to use the idea. An IRI is a unique identifier, like a web address for a thing, and a literal is a plain value, like a number or a date. The point is structural: a fact is a relationship between two things, stated in a fixed order.
Why do triples matter for search and AI?
Search engines store what they know as triples. When Google introduced its Knowledge Graph in May 2012, it launched with more than 500 million objects and "more than 3.5 billion facts about and relationships between these different objects." Each of those facts is, in effect, a triple: a thing, a relationship, and another thing or value.
That shapes how pages are read. A search engine or an AI system building an answer is trying to extract facts it can connect to things it already knows. A sentence that states a fact as subject, verb, object gives it one directly. A sentence that wraps the same fact in hedges, passive voice, and pronouns makes it guess, and a system that has to guess is less likely to rely on you.
Write the sentences that carry your key facts so a machine could turn each one into a single triple without guessing what "it" refers to.
How do you write triple-friendly sentences?
Triple-friendly sentences put a named subject first, a concrete verb second, and the fact third. The examples below use a fictional bakery; the pattern is the same for any business.
| Hard to extract | Easy to extract | The triple |
|---|---|---|
| Custom orders are something that can be taken with enough notice. | Example Bakery takes custom orders with two weeks' notice. | Example Bakery, takes custom orders with, two weeks' notice |
| It has been serving the area since the owners opened it. | Example Bakery opened in Salem, Oregon, in 2015. | Example Bakery, opened in, 2015 |
| Gluten-free options may be available depending on the day. | Example Bakery bakes gluten-free bread on Fridays. | Example Bakery, bakes gluten-free bread on, Fridays |
Three habits do most of the work. Name the subject instead of starting with "it" or "this". Prefer active verbs, so the doer is in the sentence. And put the specific fact, the number, the day, the place, inside the sentence rather than in a caption or a nearby list. You do not have to write every sentence this way, only the ones carrying facts you want understood and repeated.
How does structured data express triples?
Structured data is triples written for machines. Every property in a JSON-LD block is one fact: the thing described is the subject, the property name is the predicate, and the value is the object. Here is the bakery again, as markup:
{
"@context": "https://schema.org",
"@type": "Bakery",
"@id": "https://example.com/#bakery",
"name": "Example Bakery",
"foundingDate": "2015",
"address": { "@type": "PostalAddress", "addressLocality": "Salem", "addressRegion": "OR" },
"makesOffer": { "@type": "Offer", "itemOffered": { "@type": "Product", "name": "Wedding cakes" } }
}
Read line by line, that block says: this bakery is named Example Bakery, was founded in 2015, is located in Salem, Oregon, and offers wedding cakes. The @id gives the subject a stable identifier, so other nodes on other pages can point to the same bakery instead of describing a new one. Google defines structured data as "a standardized format for providing information about a page and classifying the page content," and requires that it match what the page visibly says. The sentences and the markup should state the same triples.
What mistakes hide the facts?
- Passive voice that drops the doer: "Wedding cakes are made to order" never says by whom.
- Pronoun chains: three sentences in a row starting with "it" leave a quoted sentence with no subject.
- Hedging on facts you know: "may", "might", and "depending on factors" turn a fact into a maybe.
- Facts only in images or lists: a price inside a graphic, or a bare bullet with no subject, has nothing to attach to.
- Markup that disagrees with the page: structured data claiming something the visible text does not say.
How do you check your pages?
Start with the facts that matter most, like what you offer, where, for whom, and at what price, and find the sentence on the page that states each one. If you cannot point to a single sentence that states a fact plainly, write one.
For a faster read of a whole page, the free Entity SEO Analyzer reports the share of declarative sentences, the share in passive voice, hedging, and how many sections open with a direct answer, alongside whether the title, headings, and body agree on one subject. It is a set of rule-based heuristics, not a model of how any engine parses text, so treat its numbers as editing prompts. Disclosure: writing and marking up pages this way is part of what I sell, so weigh the source.
Sources & further reading
Topics & entities in this article
Frequently asked questions
One fact in three parts: a subject, a relationship, and an object. "Example Bakery offers wedding cakes" is a triple: Example Bakery, offers, wedding cakes.
Not everywhere. Write the sentences that carry your key facts, like services, locations, prices, and definitions, so each states one fact with a named subject. The rest of the page can read naturally.
Schema markup is triples written for machines. In a JSON-LD block, the thing described is the subject, each property is a predicate, and each value is an object.
The term comes from the W3C's RDF standard, which defines a triple as a subject, a predicate, and an object, conventionally written in that order.
Google's Knowledge Graph stores facts about entities and the relationships between them, which is the triple structure. Google launched it in 2012 with more than 3.5 billion such facts and relationships.
Related service
Topical Authority Mapping
Topical authority mapping structures your entire topic space around entities. The map defines every pillar, cluster, and gap, so your site covers the subject comprehensively and search engines treat you as the authority.