Organization schema on 27 homepages: 12 name themselves, 2 point nowhere

FIELD TEST · 2026-08-17 · 30 sites · single snapshot

Sample and method: the same 30 sites from our robots.txt and sitemap surveys. One homepage request each, browser user agent, 2026-08-17. Three refused us, so counts are out of 27. The same crawl feeds four other posts in this batch.

Organization schema is how a page states, in machine-readable form, which company it belongs to. Sixteen of 27 homepages carry any JSON-LD at all. Fourteen of those sixteen name themselves as some kind of organization, and thirteen list sameAs links out to their other profiles. The two that declare an organization and then give it nowhere to point are both publishers.

What the 27 homepages declare

The interesting split is not who has structured data — that question we answered in an earlier survey. It is what the ones who bothered actually said about themselves.

Zero broken blocks across 16 sites is worth stating plainly. Whatever else goes wrong with structured data, malformed JSON is not the common failure at this level of the web.

How we measured it

Rules fixed before the crawl, unchanged after.

1. Pull every <script type="application/ld+json"> block from the delivered HTML.
2. Parse each one, then walk into @graph, mainEntity and itemListElement so nested nodes get counted too.
3. Collect every @type value found, whether it came back as a string or an array.
4. Separately, look for the literal string "sameAs" anywhere in the HTML.

That last rule is deliberately loose, and it has a consequence we will come back to.

Which types actually show up

Eleven distinct types across the sixteen sites. The tail is short — most of the vocabulary nobody uses on a homepage.

Netlify's single block is the densest: six types in one graph, including a BreadcrumbList on the homepage. Vercel's is the sparsest of the sixteen — one SoftwareApplication and no organization node, so the page describes a product without saying who ships it.

Declaring an organization and giving it nowhere to point

A sameAs array is the list of other places the same entity lives — the company's own profiles elsewhere. It is what lets an engine connect a page to an entity it already knows about rather than treating it as a new unknown.

Two sites declare Organization and ship no sameAs at all.

Both are large publishers whose entities are certainly already known to search engines, so the practical cost to them is probably small. For a site nobody has heard of, the same omission is the whole point of the markup going missing.

Where our own rule was too loose

BBC is the case that shows it. Its homepage carries one JSON-LD block typed only WebPage — no organization node anywhere in it — yet our sameAs check fired, because we searched the raw HTML rather than the parsed graph. The string is there; it is not attached to an organization we could see.

We are reporting BBC inside the thirteen because that is what our stated rule produced, and changing the rule after seeing the result is how surveys turn into arguments. But treat that thirteen as an upper bound.

The same looseness cuts the other way on typing. Our strict check counted only Organization and Corporation, so The Verge and The New York Times — both using NewsMediaOrganization, which is a subtype of Organization — fell outside the twelve. That is why the table above lists them on their own row instead of quietly folding them in.

What this does not tell you

We do not know whether any engine read these declarations, or what it did with them. Structured data is an input, and the output — whether a model describes your company correctly — is not observable from a page fetch. Anyone who tells you a percentage for that is measuring something else.

We also only read homepages. A site can carry a rich Organization block on an about page and nothing on the homepage, and this method would score it as bare.

What your own organization schema should carry

Four rules, and the two negative ones matter more than the positive ones because they are the failures we actually found in this sample.

• One Organization node with your legal name, logo, and a sameAs array pointing at the profiles you control.
• Keep the name string identical everywhere it appears — the markup, the page text, the title tag.
• Do not declare Organization with no sameAs. It names an entity and gives an engine no way to confirm it is the one it already knows.
• Do not point sameAs at profiles you do not control. It is a claim about identity, not a link list.

For the markup itself, our chapter on schema markup for AI search has pasteable JSON-LD for four page types. The prior survey of how many homepages carry any structured data at all is in most homepages carry no structured data. Checking this across every page rather than one at a time is what QueryWin is being built for.

Common questions

Is Corporation better than Organization?

Only one site here used it. Corporation is a subtype, so it is more specific and equally valid. Specificity helps when it is accurate and hurts when it is a guess.

Does sameAs need to include every profile?

No. It needs to include profiles you control and that carry the same identity. A short accurate list beats a long aspirational one.

Why did 11 sites have no JSON-LD at all?

We did not ask them. Several are developer tools whose homepages are heavily client-rendered, and this method only sees delivered HTML — so some of that eleven may be shipping markup we could not observe.

Does organization schema help with AI citations?

Untested here. What it verifiably does is give an engine a name, a logo and a set of confirmable profiles in one place instead of asking it to infer all three.