Structured Data for AI Search (Schema That Works)
The JSON-LD schema types LLMs and AI Overviews actually use — Article, FAQ, HowTo, Organization, Product, and more.
Why AI Engines Care About Schema
Structured data gives machines an unambiguous, extractable representation of your content. When an LLM answers 'What is X?' with a citation, it prefers sources where the answer is trivially extractable.
Google confirmed multiple times that structured data helps understanding — and therefore inclusion in Rich Results and AI Overviews.
The Schema Types That Matter
Article — every long-form post. Include headline, datePublished, dateModified, author, publisher, image.
Organization — homepage. Include name, url, logo, sameAs (LinkedIn, GitHub, Twitter, Wikipedia).
Person — About page for personal brands. Include jobTitle, worksFor, sameAs.
FAQPage — one per long-form article with real Q&A.
HowTo — procedural, step-by-step content.
Product — every ecommerce product with Offer, AggregateRating, Review.
BreadcrumbList — every deep page. Helps navigation snippets and entity resolution.
WebSite with SearchAction — homepage. Enables sitelinks search box.
Schema types by page role
| Page | Primary schema | Optional add-ons |
|---|---|---|
| Homepage | Organization, WebSite | SearchAction, sameAs |
| Article | Article, FAQPage | BreadcrumbList, HowTo |
| About | Person or Organization | sameAs, alumniOf, worksFor |
| Product (PDP) | Product, Offer | AggregateRating, Review, BreadcrumbList |
| Category | CollectionPage | BreadcrumbList, ItemList |
| Contact | LocalBusiness (if applicable) | OpeningHoursSpecification, PostalAddress |
Format: JSON-LD, Not Microdata
JSON-LD in the <head> or before </body> is the modern standard. It's decoupled from the visible HTML, easier to maintain, and preferred by Google.
Microdata (inline HTML attributes) and RDFa still work but are legacy patterns you should migrate away from.
Common Mistakes
Marking up content that isn't visible on the page — a policy violation.
FAQPage with fake or generic Q&As stuffed for schema — a manual action risk.
Missing required fields — Rich Results Test flags these.
Duplicated schemas from CMS + plugin + hand-coded — deduplicate to one canonical block per type.
Validation Workflow
1) Google's Rich Results Test for eligible-rich-result checks.
2) Schema.org validator for structural correctness.
3) Search Console's Enhancements report for site-wide health.
Frequently Asked Questions
Does FAQ schema still show rich results?+
Google reduced FAQ rich results in 2023 to trusted government and health domains for the main SERP, but the schema is still valuable for AI Overviews, Perplexity, and Bing.
Is HowTo schema still useful?+
Yes for extraction, less for rich results (also restricted in Google SERP). Ship it for procedural content — LLMs use it heavily.
What about Product schema in AI Overviews?+
Very useful. Correct Product + Offer + AggregateRating markup increases Merchant Center coverage and appears in AI product answers.
Can I use multiple schema types on one page?+
Yes — combine them in a single JSON-LD @graph, or ship separate blocks. Both are valid.
Does schema improve rankings directly?+
Not as a direct ranking factor. But it improves crawler understanding, unlocks rich results, and increases AI-answer citations — all of which lift click-through and visibility.