Structured data and AI answers
Published Updated
Structured data is code that labels the facts on your page in a form machines read reliably: who you are, what you sell, what an article is about. It can make your pages easier to interpret, but no major assistant promises to recommend a business because of it. Use it to state true facts clearly, and do not expect it to do the work of good content.
Think of it as clean labelling on a well-written page, not as a lever.
What is structured data?
It is a standard way to describe a page's content using a shared vocabulary, most commonly schema.org. The recommended format for most search engines is JSON-LD, a block of data placed in the page that is separate from the visible text.
A tiny example for a company:
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Example Co",
"url": "https://example.com",
"description": "Software that does one clear thing for agencies.",
"sameAs": ["https://www.linkedin.com/company/example"]
}That block says: this is an organisation, here is its name, address on the web, a plain description and where else it appears. A machine does not have to infer any of it from layout.
Does it affect AI answers?
Be careful here. Search engines have documented how they use structured data, for example to show rich results. For AI assistants, the public picture is thinner. Some assistants rely on search results, and those can use structured data. Beyond that, how any assistant weighs it is not something to guess at.
So hold two thoughts together:
- Structured data can make the facts on your site less ambiguous for any system that reads it.
- There is no known setting, schema type or trick that makes an assistant recommend you.
The honest claim is modest: it reduces room for misreading. That is worth having, especially for facts that are easy to get wrong, such as your name, what you sell and your prices.
Which types should a business mark up?
Start with the ones that describe real, visible content.
| Type | Use it for |
|---|---|
| Organization | Your company: name, URL, logo, description, profiles elsewhere |
| WebSite | The site itself, including its name |
| Product or SoftwareApplication | A product page: name, description, offers, category |
| Article or BlogPosting | Guides and posts: headline, author, dates |
| FAQPage | A page with a real, visible list of questions and answers |
| BreadcrumbList | The page's place in your site structure |
Skip anything you cannot fill truthfully. An empty or invented field is worse than a missing one.
What rules keep it honest?
- Mark up what readers can see. If the markup says something the page does not, it conflicts with the content and may be ignored.
- Keep it current. A price in markup that differs from the price on the page is an error you created yourself.
- Be specific. A clear description beats a keyword list.
- Do not mark up reviews you did not earn, or ratings that are not shown to visitors.
- Use one consistent name for your company across the site and profiles. It helps with the confusion that causes wrong claims in AI answers.
How do you test it?
Use the validators the search engines provide and a plain read-through.
- Run the page through a structured data testing tool from a major search engine, and fix errors and warnings that apply to you.
- View the page source and check the JSON-LD is present in the delivered HTML, not added only after heavy scripts run.
- Compare the markup with the visible page. Every fact in one should be in the other.
- Re-test after template changes. Site updates are the usual way markup breaks without anyone noticing.
Where does it sit among the other levers?
Structured data is one of several things that make your site easy to read. Compare it with the rest.
- Crawl access: if crawlers are blocked, markup is never seen. See AI crawlers and robots.txt.
- A clear homepage: plain language beats labels on a confusing page. See why your homepage confuses AI assistants.
- Third-party presence: many recommendations rest on what other pages say. See the kinds of pages AI assistants cite most.
- llms.txt: a different, optional file. See llms.txt explained.
If you can only do one thing this week, make the visible page clearer. Add the markup when the page is right.
How do you know whether it helped?
Treat it as you would any other change. Record how assistants answer your key questions, and which pages they cite, before the change. Repeat after some weeks. Only call a difference real if it is bigger than the normal run-to-run variation. How to track AI visibility over time covers the routine.
Who should own the markup?
Someone on your team should be responsible for it. Markup is often added once by a developer or plugin and then forgotten while prices, names and products change. Put it on a short review list alongside your pricing page. When a fact changes in one place, check the other. A small amount of well-kept markup beats a large amount nobody maintains.
To see how a site looks to AI assistants today, try the Citewake checker. For the way Citewake reads pages and records what it finds, see the methodology.
Questions
- Does structured data make AI assistants recommend me?
- There is no documented guarantee. It helps machines read facts about your pages and business more reliably, which is useful, but it is not a recommendation switch.
- What format should I use for structured data?
- JSON-LD using schema.org vocabulary is the format most search engines recommend. It sits in a script block and keeps markup separate from visible content.
- Which types are most useful for a business site?
- Organization, WebSite, Product or SoftwareApplication where relevant, Article for content, and FAQPage only where a real FAQ is visible on the page.
- Can wrong structured data cause problems?
- Yes. Markup that contradicts the visible page, or marks up things readers cannot see, can be ignored or penalised by search engines and creates conflicting facts.