Complete Organization schema: the machine-readable ID card of your brand

On your site there is a technical record describing your company in a language that AIs read directly — and it probably has almost all of its fields empty. This is not a detail for programmers: the models count how much information they can cross-reference about you to decide whether they trust you enough to recommend you. Your competitor, with the same quality of services but a filled-in record, gets cited — you don't. Closing that gap takes thirty minutes, once you understand what to fill in.

Take your site, open the source code of the homepage (right click → “View page source” or Ctrl+U), search with Ctrl+F for the word “Organization”. Want to bet it isn’t there, or that just two scattered fields show up — `name` and `url` — and nothing else?

If that’s the case, you’re giving ChatGPT, Perplexity, Gemini and Claude an ID card with first name, last name, and then the rest left blank. The AI engine reads you, but with less confidence than a competitor of yours who has filled in the 25 fields that the Organization schema makes available. And if it has to choose who to cite in the answer to a user asking “best dental practice in Catania”, with equal content it picks the one with the verifiable identity.

I’ll explain why the number of fields in the Organization schema is a trust signal for the AI, how to check them in three minutes, and what happened when I compared two comparable Sicilian dental practices — one with a complete schema, one with a skeletal schema.

What the AI says when it reads your Organization schema

In the world of structured web search, the principle is simple: schema.org is the shared vocabulary with which engines translate the content of a page into an entity with attributes. Guha, Brickley and Macbeth — the founding fathers of schema.org, formerly at Google, Bing, Yahoo, Yandex — documented in the paper Schema.org: Evolution of Structured Data on the Web (Communications of the ACM 2016) how this vocabulary became the de facto standard for making engines understand “what this thing is” instead of “which words are written here”.

The `Organization` type has over 25 standard properties: `name`, `legalName`, `url`, `logo`, `description`, `foundingDate`, `founder`, `address`, `contactPoint`, `email`, `telephone`, `sameAs`, `numberOfEmployees`, `areaServed`, `knowsAbout`, `identifier`, `vatID`, `taxID`, and others.

It follows from this that every filled-in field is one more piece of identity. The AI doesn’t read “Organization schema present/absent”, it reads how many verifiable attributes that entity has. An entity with 18 filled-in fields, including a `sameAs` pointing to LinkedIn, Wikidata, Google Business Profile and a professional register, is a “strong” identity. An entity with 3 fields is a “weak” identity: it exists, but the AI has no footholds to cross-reference and validate.

The Web Data Commons team (2023 paper on the schema.org tables corpus) quantified the growth: tens of millions of domains publish schema.org data, and AI engines — which train on web crawls — use it as a grounding signal when they have to decide whether “this organization really exists” and “what exactly it does”.

Why it sits upstream of everything else in AI visibility

In the previous articles of this series on the knowledge graph I explained how AI engines build a graph of entities starting from the web. The Organization schema is the entry point of YOUR entity into that graph. If it’s skeletal, you enter the graph as a weak node, with few connections. If it’s complete, you enter as a robust node, connected to Wikidata, to LinkedIn, to your Google Business profile, to the professional register.

It’s the same logic as the vector space of embeddings I told you about in P1: the more coherent signals you give, the “denser” your vector is in the model’s representation space. And it’s the technical prerequisite of E-E-A-T for AI — you declare expertise in the content, but authority and trustworthiness also pass through here: through the engine’s ability to verify that you are who you say you are.

To put it another way: the content makes you say things, the Organization schema makes you recognizable as an entity. Without the second, the first carries less weight in your visibility within AI answers.

The test you can run on your site in three minutes

Here is the operational check, step by step:

1. Open Google’s Rich Results Test, paste the homepage URL, click “Test URL”.
2. In the right-hand column, under “Detected items”, look for “Organization”. If it’s not there, the problem is upstream: you don’t have an Organization schema. If it is there, click on it.
3. Look at the list of filled-in properties. Count them.

The practical threshold I use with clients:

• Fewer than 8 fields: skeletal identity. The AI sees you, but with low confidence.
• Between 8 and 15 fields: acceptable identity. You’re above the Italian SME average.
• Over 15 fields, with `sameAs` pointing to at least 4 verifiable external profiles: robust identity. You’re at the level of the brands that the AI cites with ease.

It’s an entry-level test, of course — the real analysis of the knowledge graph requires professional tools and a precise comparison with the competitors in your sector. But to understand whether you’re at the “ID card filled in” or “document with empty fields” level, three minutes are enough.

The A/B comparison I ran on two dental practices in Catania

I took two comparable Sicilian dental practices: both in Catania, both with 3-4 chairs, both specialized in implantology and orthodontics, both with a WordPress site and an active blog for at least 3 years. Let’s call them Practice A and Practice B (anonymized).

Practice A had an Organization schema with 4 fields: `name`, `url`, `logo`, `telephone`. That’s it.

Practice B had an Organization schema with 17 fields: all of A’s plus `description`, `foundingDate`, `founder` (with the name of the owner dentist and a link to their `Person` page), a complete `address` (street, postal code, city, Sicily region), `contactPoint` with email and hours, `areaServed` (Catania and its province), `sameAs` with 6 links (owner’s LinkedIn, practice Facebook page, Google Business profile, the dentist’s Wikidata, Order of Physicians of Catania, Instagram).

I queried ChatGPT, Perplexity and Gemini with 12 queries like “best implantology practice Catania”, “children’s orthodontics practice Catania center”, “dentist specialized in aesthetic veneers eastern Sicily”, and other sector variations.

Result on the sample (indicative test, not a controlled study): Practice B was cited in 8 answers out of 12, Practice A in 1. The content of the two blogs was comparable in volume and perceived quality. The most evident structural difference was the schema.

Honest limitation of the test: 12 queries are not a statistical study, and other variables (backlinks, mentions on health portals, Google reviews) play a role. But the pattern is consistent with what I see in the A/B comparisons I run regularly: with equal content, the node with a robust Organization schema wins on visibility in AI answers.

The mistakes I see most often

When I open the Rich Results Test on clients’ sites, four patterns recur:

• Only `name` and `url`: the schema is there because the WordPress theme generates it by default, but no one ever extended it. This is 60% of cases.
• `sameAs` empty or with a single link: only Facebook is cited, or nothing. This way the AI can’t cross-reference your identity with authoritative external sources.
• `address` with only the city, without street and postal code: for a local business this is a weak identity signal. The AI doesn’t understand whether you’re really in Catania or whether you “cover” Catania.
• `founder` missing: in professional practices and small businesses the owner is the trust entity. Not declaring it in the schema is equivalent to not introducing yourself.

What to do concretely this week

Fill in, in this order:

• `name`, `legalName` (if different), `url`, `logo` (high-quality PNG/SVG image URL)
• `description` (1-2 sentences, the same ones you’d use in a press release)
• `foundingDate`, `founder` (owner’s name, with a link to a dedicated `Person` page if it exists)
• complete `address`: street, postal code, city, region, country
• `contactPoint`: email, telephone, hours, languages spoken
• `areaServed`: cities and province or regions you really serve
• `sameAs`: insert ALL the official profiles — company LinkedIn, owner’s LinkedIn, Facebook, Instagram, Google Business Profile, Wikidata (if you have one), professional register, authoritative industry portals
• `numberOfEmployees`, `vatID` (VAT number): where relevant

After each change, run the Rich Results Test again to check that Google reads it without errors. Then compare the number of filled-in fields with the 3-5 competitors that the AI cites in your sector when you run the test queries: if they’re at 15 and you’re at 6, you know where to act.

Where all this takes you

A complete Organization schema is not a magic factor, and on its own it won’t get you into AI answers. But it’s the machine-readable passport of your entity: without it, every other effort on content carries less weight. With it, the content you already have starts to carry more weight because the engine knows with greater confidence who is publishing it.

In the next articles of this series we’ll see how to extend the same logic beyond Organization: the `Person` schema for the owner (your personal entity in the graph), the `LocalBusiness` schema for geo-local specifics, and the connection with Wikidata that closes the loop and makes your entity verifiable outside your own site. These are the next steps to consolidate visibility in AI answers, starting from the foundations.