Schema Markup for AI Search: The Types That Actually Matter
For AI visibility the types that matter most are Organization (who you are, verifiably), Service or Product (what you sell), FAQPage (questions you answer) and Article (what you publish). Connected into one graph with stable identifiers, they tell an engine what your business is — which is the prerequisite for being named.
Structured data is often treated as a rich-results tactic, which undersells it. For answer engines its value is different: it is the most unambiguous statement available about what a business is, what it sells and what it claims — in a form that needs no interpretation and can be checked against other sources.
- Schema that contradicts the page is worse than no schema.
- Connect nodes with @id rather than emitting disconnected fragments.
- Organization is the foundation — it is the entity everything else attaches to.
- Two plugins both emitting schema is a common and damaging mistake.
The types worth implementing
| Type | Where | Why it matters for AI |
|---|---|---|
| Organization | Sitewide | Establishes the entity. Include legalName, company number as identifier, address, sameAs |
| Service | Service pages | States what you sell and where, linked to the provider |
| Product / SoftwareApplication | Product pages | Category, audience, features, offers |
| Offer | Pricing | A machine-readable price — one of the few things reliably quoted |
| FAQPage | Pages with real questions | Explicit question-answer pairs, directly usable |
| Article | Journal posts | Author, dates, language — the provenance an engine weighs |
| BreadcrumbList | All interior pages | Site structure, cheap to add |
Connect it into one graph
The common mistake is emitting several disconnected blocks — an Organization here, an Article there, with nothing linking them. Use a single @graph with stable @id values, so the Article’s publisher points at the Organization node and the Service’s provider does too. That turns a pile of assertions into a description of one business, which is what an engine needs to resolve who you are.
Mistakes that make it counterproductive
- Claiming things the page does not say. FAQ markup for questions that are not on the page is the most common version, and it is a policy violation as well as a credibility problem.
- Two sources emitting schema. A theme and a plugin both producing Organization data means contradictory claims about the same entity.
- Review markup on reviews you do not have. Treat this as a line not to cross.
- Facts that disagree with your other sources. A different company name in schema than on Companies House undermines the corroboration the markup was meant to provide.
- Leaving it to drift. Schema written once and never updated eventually describes a business that no longer exists.
Frequently asked questions
Does schema improve rankings?
Not directly as a ranking factor. It affects how results are displayed and how clearly engines understand a page, both of which affect outcomes. For AI visibility specifically it does more work, because it is unambiguous where prose is interpretable.
JSON-LD or microdata?
JSON-LD. Google recommends it, it keeps markup separate from presentation, and it is far easier to maintain than attributes scattered through templates.
Is FAQ schema still worth it?
Yes, though its visible rich-result treatment has been reduced. It remains the clearest machine-readable statement of what a page answers, and answer engines read it — the value shifted rather than disappeared.
How do we check ours is right?
Google’s Rich Results Test and the Schema.org validator for syntax, then read the output yourself and confirm every claim appears on the page. Validators check structure, not honesty.
Schema implementation is part of AI SEO, and the visibility audit validates what you currently emit.
