Schema markup gets sold as an AI visibility lever, which overstates it in a way worth correcting. Adding structured data does not cause a system to cite you. What it does is make your content unambiguous to a machine, which is a precondition for being considered rather than a cause of being selected.
The distinction matters, because it tells you how much effort schema deserves. Enough to be legible. Not enough to be a strategy.
Why it helps at all
An AI system building an answer has to work out what a page is about, who wrote it, when it was published and whether the organisation behind it is a recognisable entity. Unstructured prose forces inference. Structured data states it.
Pages carrying robust markup do appear to earn citations at higher rates, and the plausible mechanism is straightforward: they are cheaper to interpret and safer to attribute. Nothing mystical.
The types that carry weight
Organization is the foundation and the one most sites do badly. It establishes you as an entity: name, logo, address, contact points, and sameAs links to your verified profiles elsewhere. Entity clarity is repeatedly identified as a factor in citation selection, and this is where it lives.
Article with a real author, including a person rather than a site name, a publish date and a modified date. Author signals feed directly into whether a source looks accountable.
FAQPage for genuine question and answer content. This is the type most directly aligned with how AI answers are assembled, because it presents a discrete question with a self contained answer.
LocalBusiness if you serve a location. Local queries increasingly trigger AI surfaces, and markup here influences whether you are named.
Product and Review for anything commercial, including shipping and returns, which assistants surface in comparisons.
The official Google structured data documentation is the reference worth using directly rather than through commentary.
The types that are mostly theatre
Marking up everything is not a strategy, and some of it is actively counterproductive.
- HowTo lost its rich result treatment and adds little now unless the content genuinely is a procedure.
- Speakable remains limited in practice.
- Deeply nested WebPage and BreadcrumbList chains that restate what is already obvious add weight without adding meaning.
- FAQPage on content that is not questions. Inventing questions to qualify for markup is exactly the sort of manipulation that gets policed.
The part schema cannot do
This is the honest limit. Schema describes content. It does not improve it.
A page that restates general knowledge, perfectly marked up, gives an AI system nothing to cite that it could not generate itself. Citation goes to pages carrying something specific: original data, a documented process, actual pricing, a named practitioner opinion, a result with numbers attached.
If your page has none of that, no amount of structured data creates it. Markup makes a good page findable. It does not make a thin page worth quoting.
What to do this week
- Validate what you currently output with the Schema.org validator. Most WordPress sites emit schema from a plugin, a theme, and sometimes both, producing conflicting or duplicate blocks.
- Fix Organization first, with sameAs links to your real profiles.
- Add real author markup to articles, pointing at a person with a genuine bio page.
- Add FAQPage only where actual questions and answers exist.
- Then stop, and spend the remaining effort on making the content worth citing.
Our guide to getting cited in AI answers covers the content side, and reading the Search Console AI report covers measuring whether any of it worked.
Frequently asked questions
Does schema markup help you appear in AI Overviews?
It helps indirectly. Structured data removes ambiguity about what a page is, who wrote it and when, which makes the page cheaper to interpret and safer to attribute. It does not cause selection, and pages with no distinctive content will not be cited regardless of markup quality.
Which schema types matter most for AI search?
Organization with sameAs links, Article with a real named author and dates, FAQPage where genuine questions exist, LocalBusiness if you serve a location, and Product with shipping and returns for commercial pages. These cover entity clarity, accountability and answerability.
Is FAQ schema still worth adding?
Yes, where the content genuinely is questions and answers, because that format matches how AI answers are assembled from discrete self contained responses. Inventing questions purely to qualify for the markup is manipulation and is the kind of thing spam policies address.
Can too much schema hurt you?
Conflicting or duplicated markup causes real problems, and many WordPress sites emit schema from both a plugin and a theme simultaneously. Marking up content inaccurately to qualify for a feature is a policy violation. Excess markup that is merely redundant mostly wastes effort rather than causing harm.
How do I check my schema is working?
Run key pages through Google Rich Results Test and the Schema.org validator, and check the Enhancements reports in Search Console for errors. Look specifically for duplicate blocks, since a plugin and a theme both outputting Organization data is a common and quiet source of conflict.