sameAs vs knowsAbout: which Schema.org property when

TL;DR

• Use sameAs for identity: URLs that prove "this entity is the same entity" on another authoritative source (Wikipedia, Wikidata, LinkedIn, Crunchbase).
• Use knowsAbout for topical expertise: 5–15 topics a Person or Organization has demonstrable authority in.
• They are not interchangeable. sameAs disambiguates who you are; knowsAbout describes what you know.
• For AI citation work, use both together on Person and Organization nodes, and reference each knowsAbout topic to a Wikidata @id plus a Wikipedia sameAs.
• Most common misuse: dumping topical URLs into sameAs or listing 40+ topics in knowsAbout. Both dilute the signal engines were built to read.

Schema.org gives you two properties that look superficially similar but do very different work in an entity graph. sameAs links your entity to a reference page that unambiguously identifies it. knowsAbout declares subjects a Person or Organization is knowledgeable in. Confusing them is one of the most common structured-data errors I see when auditing sites for AI visibility, and it directly affects whether ChatGPT, Perplexity, and Google's AI Overviews treat you as a credible source.

What sameAs is actually for

sameAs takes a URL that resolves to a page representing the same real-world entity. Wikipedia is the canonical example: if there's a Wikipedia page for your company or founder, that URL belongs in sameAs. So does the matching Wikidata Q-item, an official LinkedIn profile, a Crunchbase page, a GitHub organization, a verified X account, an ORCID for an academic author, or an IMDb page for a filmmaker.

The rule is identity, not topic. A link to a Wikipedia article about marketing analytics does not belong in your CMO's sameAs — she is not the concept of marketing analytics. A link to her Wikipedia biography does.

{
  "@context": "https://schema.org",
  "@type": "Person",
  "name": "Jane Okafor",
  "jobTitle": "Chief Data Officer",
  "worksFor": { "@type": "Organization", "name": "Northwind Analytics" },
  "sameAs": [
    "https://en.wikipedia.org/wiki/Jane_Okafor",
    "https://www.wikidata.org/wiki/Q123456789",
    "https://www.linkedin.com/in/janeokafor",
    "https://orcid.org/0000-0002-1825-0097",
    "https://scholar.google.com/citations?user=abc123"
  ]
}

Every URL above answers "is this the same person?" — which is exactly what an LLM's entity resolver needs to fuse your site's mention of "Jane Okafor" with the Jane Okafor already sitting in Google's Knowledge Graph.

What knowsAbout is actually for

knowsAbout is a Person/Organization property describing subjects the entity is expert in. It's how you tell engines that Jane doesn't just exist — she has demonstrable authority in specific topics that your content covers.

Two rules matter here:

1. Keep the list tight. Five to fifteen topics is the practical range. Below five and you're under-describing a real expert; above fifteen and every topic gets weaker because the entity looks unfocused. Treat it like a résumé, not a tag cloud.
2. Ground each topic in the Knowledge Graph. A bare string works, but a nested Thing with @id pointing to Wikidata and sameAs pointing to Wikipedia is dramatically stronger. Wikidata and Wikipedia are both reference sources for Google's Knowledge Graph, and they don't perfectly overlap — some concepts exist in one but not the other. Linking both maximizes disambiguation.

{
  "@context": "https://schema.org",
  "@type": "Person",
  "name": "Jane Okafor",
  "knowsAbout": [
    {
      "@type": "Thing",
      "name": "Marketing analytics",
      "@id": "https://www.wikidata.org/wiki/Q6765423",
      "sameAs": "https://en.wikipedia.org/wiki/Marketing_analytics"
    },
    {
      "@type": "Thing",
      "name": "Attribution modeling",
      "@id": "https://www.wikidata.org/wiki/Q4818842",
      "sameAs": "https://en.wikipedia.org/wiki/Attribution_(marketing)"
    },
    {
      "@type": "Thing",
      "name": "Bayesian inference",
      "@id": "https://www.wikidata.org/wiki/Q812535",
      "sameAs": "https://en.wikipedia.org/wiki/Bayesian_inference"
    }
  ]
}

Notice how each topic is something a reader would actively search expertise for. "Business" or "digital" wouldn't clear that bar; "attribution modeling" does.

Common misuse patterns

Three mistakes come up repeatedly in audits:

• sameAs as a topic dump. Marketers paste Wikipedia URLs for industries and skills into sameAs, thinking it will boost topical relevance. It doesn't — it corrupts the identity signal. Engines that trust sameAs for entity resolution will either ignore the property or downweight the entire node.
• knowsAbout as a keyword bag. Long lists of 30–60 topics, most of them tangential, tell an LLM the entity has no real specialty. Prune ruthlessly.
• String-only knowsAbout for competitive niches. "knowsAbout": ["SEO", "content marketing", "PPC"] is valid schema but weak. Every consultant on earth claims those strings. Wikidata IDs give you disambiguation your competitors don't have.

A related mistake: putting knowsAbout on the wrong type. It's defined for Person and Organization. Don't attach it to WebPage, Article, or Product.

Using both together for GEO

For AI citation work, the two properties are complements, not alternatives. When an AI engine evaluates whether an author is credible on a topic, it looks at knowsAbout alongside jobTitle, worksFor, and alumniOf to build an expertise profile — then uses sameAs to verify that the profile matches a known entity in its graph. Skip sameAs and the expertise claim floats untethered; skip knowsAbout and the resolved entity has no declared subject authority.

Practical implementation order:

1. On every Person (author) node, add sameAs for identity (LinkedIn, Wikipedia if applicable, ORCID, personal site, verified socials).
2. Add knowsAbout with 5–15 nested Thing objects, each with @id (Wikidata) and sameAs (Wikipedia).
3. On the Organization node, do the same — sameAs for Crunchbase, LinkedIn, Wikipedia, Wikidata; knowsAbout for the company's core practice areas.
4. Make sure the topics in knowsAbout actually match the articles the person publishes. An AI engine will notice the mismatch if the author "knows about" quantum computing but only writes about email marketing.

Done well, this gives ChatGPT, Perplexity, Gemini, and AI Overviews the two things they need most: a resolvable identity and a scoped domain of expertise.

FAQ

Can I put topic URLs in sameAs to boost topical relevance?

No. sameAs is strictly for URLs that identify the same entity. Putting topic pages there will either be ignored or weaken the identity signal engines rely on. Use knowsAbout for topics.

How many sameAs links should an entity have?

There's no hard cap, but every URL should be authoritative and unambiguous. Five to ten high-quality identity references (Wikipedia, Wikidata, LinkedIn, Crunchbase, ORCID, verified socials) beats twenty weak ones.

Does knowsAbout help if my entity isn't in Wikidata yet?

Yes — engines still read the property, and grounding topics in Wikidata/Wikipedia URLs helps even when the entity itself has no Wikidata item. Getting the entity into Wikidata is a separate, high-leverage project worth prioritizing.

Sources

• Schema.org: sameAs
• Schema.org: knowsAbout
• Schema.org: Person
• Schema.org: Organization
• Wikidata
• Google: Introduction to structured data markup