Schema markup is your website’s resume.

Schema markup is your website's resume, and for AI visibility that means both less and more than the hype suggests. It will not get you picked by ChatGPT or Google's AI on its own, and anyone selling it as an AI citation hack is overselling. What it does is make sure the machines screening you read the right name, the right credentials and the right references. A missing, vague or dishonest resume gets you screened out before the interview.

That is the argument. The rest of this essay explains what schema actually is, what the evidence says it does and does not do for AI visibility, and a five-line framework I use to decide which markup is worth the developer's time.

What is schema markup, in plain language?

Schema markup is a block of structured code on a web page that tells machines, in a fixed vocabulary, what the page is about. The vocabulary comes from Schema.org, a shared standard backed by the major search engines and a basic part of search engine optimisation. The most common format is JSON-LD, a small script that sits in the page and is invisible to human visitors.

Your page says, in prose, "Dr Sara Khan is a consultant cardiologist at our Dubai clinic." A human reads that easily. A machine has to guess: is Sara Khan a person, a clinic, a brand? Is "cardiologist" a job title or a department? Schema removes the guessing:

{
  "@context": "https://schema.org",
  "@type": "Physician",
  "name": "Dr Sara Khan",
  "medicalSpecialty": "Cardiovascular",
  "worksFor": { "@type": "MedicalClinic", "name": "Example Clinic Dubai" },
  "sameAs": ["https://www.linkedin.com/in/example-profile"]
}

An illustrative example with a fictional doctor and clinic.

Google's own documentation describes structured data as a way of "providing explicit clues about the meaning of a page". That phrase is the whole idea. Schema does not make your content better. It makes your content unambiguous.

Schema is not content. It is a statement about your content, in a language machines do not have to interpret.

Why "resume" is the right way to think about it

A resume does not do the job for you. It does something narrower and still decisive: it gets you correctly understood by someone screening hundreds of candidates in seconds. Name, role, credentials, references, relevant work. Structured, scannable, verifiable.

That is exactly the job schema does for a website, and the parallel holds further than it first appears:

Why "new"? Because the screener changed. For twenty years the main reader of your structured data was Google's results page, and the payoff was visible: star ratings, FAQ dropdowns, recipe cards. Now the reader is a stack of systems (search indexes, knowledge graphs, retrieval layers) that feed answers in AI Overviews, ChatGPT, Copilot, Perplexity and Gemini. The payoff is less visible. The screening is more important.

The old resume won you a prettier listing. The new one decides whether the machine knows who you are at all.

Does schema markup help AI visibility? What the evidence says

Not directly, and I think marketers should hear that before they hear anything else. Here is the opposing view, stated fairly: AI engines prefer clean, structured facts; schema is the cleanest structured fact on the page; therefore more schema means more citations. It is a tidy argument. It is also the pitch behind a lot of "AI schema" tooling. The evidence does not support it.

Two caveats on that evidence. Both studies come from vendors with their own products to sell, and OtterlyAI's is a single SaaS website. Neither proves schema is useless. They prove something narrower: schema is not a citation lever you can pull and watch move.

My read: if someone promises you AI citations from a schema project, ask them which platform, which query and which measurement. You will usually get a pause.

Schema works upstream, not at the answer

So why do I still put schema near the top of every AI search readiness audit? Because the AI answer is the last step in a chain, and schema does its work earlier in that chain.

Most AI search answers are grounded in a search index: the model retrieves pages, then writes. Those indexes, and the knowledge graphs behind them, are built by crawlers that do read structured data. Schema helps those systems resolve who you are, which entity a page describes and how your pages relate to each other. The model then inherits that understanding, whether or not it ever sees your JSON-LD.

There is also one first-party confirmation. At SMX Munich in March 2025, Microsoft's Fabrice Canel confirmed that schema markup helps Microsoft's LLMs understand content for Bing and Copilot. (Search Engine Land)

Put the pieces together and the honest position is this:

This is also why I treat schema as the first of several layers, not the whole strategy. I wrote about the wider picture in AI search readiness, in plain language. Schema is the part that is mostly engineering. Authority and a distinctive editorial point of view do the rest.

The Five-Line Schema Resume

Schema.org has hundreds of types. Most websites need five things done well before they need anything else. I organise them the way a recruiter reads a resume, top to bottom.

Source: author's framework. Type names from Schema.org.

Get these five right and you have a resume. Add anything else before them and you have decoration.

Never put anything on the resume you cannot defend in the interview

Every recruiter has met the candidate whose resume says "fluent in Arabic" and whose interview says otherwise. Search engines have met that website too.

Google's structured data guidelines are clear that markup must represent the content visible on the page, and that misleading or hidden markup can lead to a manual action. (Google Search Central)

The history of FAQ schema is the cautionary tale. It was used, then overused, to grab extra space on the results page. In 2023 Google restricted FAQ rich results to well-known government and health websites. In May 2026 it stopped showing them altogether. (Search Engine Journal) Teams who added FAQ markup purely for the visual payoff lost the payoff. Teams who used it to structure real questions their customers ask still have useful, well-structured pages.

Markup written to game a feature dies with the feature. Markup written to describe the truth outlives it.

What changes in healthcare

Healthcare is where I would push hardest on schema, for a simple reason: the cost of being misunderstood is higher. A patient asking an AI assistant for a paediatric cardiologist in Abu Dhabi needs the right doctor, at the right facility, with the right specialty. A blended or outdated answer is not a lost click. It is a lost patient, or a wrong visit.

Three things I check first on any hospital or clinic website:

Two guardrails. Schema should describe services and people, never make clinical claims the visible page does not make. And healthcare advertising rules in markets like the UAE apply to what the page says; markup does not get a separate, looser standard.

In healthcare, schema is less about being found and more about not being confused with someone else.

How to audit your schema resume in an afternoon

You do not need a platform to start. You need four tabs and some honesty.

Where the resume metaphor breaks

Two places, and they are worth being honest about.

First, a human recruiter reads every resume. AI systems mostly do not read your JSON-LD directly; the evidence above suggests many cannot. Schema's influence is indirect, through the indexes and graphs those systems depend on. That makes it harder to measure, and it means you should not expect a before-and-after chart in six weeks.

Second, a resume is your main pitch. Schema is not. The things that make an AI cite you are the same things that make a person trust you: authority others grant you, and content with a point of view worth quoting. I have not seen a strong schema implementation rescue weak content, and I would be suspicious of anyone who claims they have.

Schema is necessary hygiene, not a growth strategy. Budget for it like hygiene: done properly, once, and maintained.

Frequently asked questions