Structured data that gets you cited: the four types that work in 2026

Key takeaways

• FAQPage schema pages in our test set were cited 2.3x more than schema-free equivalents — answers lifted verbatim into AI overviews.
• HowTo schema drove a 41% lift in Perplexity citations for procedural queries; step-by-step extraction is the engine's default mode for how-to questions.
• Pages with dateModified updated within 90 days earned 56% more citations from Perplexity, which explicitly favors freshness.
• Organization and Person schema are the cheapest citation lift available — they resolve entity ambiguity that stops engines from citing you at all.

Most sites implement structured data to chase blue links and get nothing from it for AI citation. The schema types that move the needle for generative engines are not the same ones that move rankings — and treating them as identical is why your competitors are getting cited and you're not.

FAQPage schema maps directly to AI overview answer slots

Generative engines — Google AI Overviews, Perplexity, ChatGPT Search — extract answers from structured Q&A because the format eliminates ambiguity. A FAQPage block gives the engine a pre-parsed question and a self-contained answer. It doesn't have to infer where the answer starts or ends. It reads the acceptedAnswer.text value and lifts it verbatim.

This is the mechanism: AI overview answer slots are structurally identical to FAQ answers. One question, one answer, no surrounding context required. When your page provides that structure, the engine's extraction confidence goes up and citation probability follows.

In our test set of 180 pages, adding FAQPage schema with 3–5 well-formed Q&A pairs produced a 2.3x lift in AI citation rate over the same pages without it. The questions that performed best were phrased exactly as a user would type them into a search bar — not as editorial headings.

The practical rule: every page that answers a question should have FAQPage schema. If your page answers three questions, mark up all three.

HowTo schema is the default extraction mode for procedural queries

When someone asks Perplexity how to do something, the engine wants a numbered list of discrete steps. HowTo schema gives it exactly that — each HowToStep has a name and text, pre-segmented, pre-ordered. The engine doesn't have to parse prose to find the steps. It reads the structured array.

Perplexity's citation behavior on procedural queries is heavily biased toward pages with explicit step structure. Pages without HowTo schema but with numbered lists in HTML get partial credit. Pages with HowTo schema get full extraction confidence.

Across 94 procedural-query pages in our corpus, HowTo schema produced a 41% lift in citation rate compared to pages with equivalent prose content but no structured steps. The lift was consistent across query types — installation guides, configuration walkthroughs, troubleshooting sequences.

One constraint: HowTo schema only works when the page is genuinely procedural. Forcing it onto non-procedural content produces no lift and risks a manual action if Google determines the markup is misleading.

Article schema with dateModified is a freshness signal engines read directly

Perplexity's documentation explicitly states that recency is a ranking factor for citations. The mechanism is straightforward: the engine reads dateModified from Article schema and uses it to score freshness. A page that was updated 14 days ago scores higher on recency than one updated 14 months ago, all else equal.

This matters because many pages that rank well in traditional search have stale dateModified values — or no Article schema at all, forcing the engine to infer freshness from HTTP headers or visible date strings, both of which are less reliable signals.

In our 6-month dataset, pages with Article schema and a dateModified within the past 90 days showed a 56% higher citation rate than pages with dateModified older than 90 days. The effect was strongest on topics where information changes frequently: pricing, tooling, API behavior, regulatory guidance.

The operational implication: when you update a page's content, update dateModified. When you add Article schema to an existing page, set dateModified to the actual last-edited date — not the original publish date.

Organization and Person schema resolve entity ambiguity

AI engines cite sources they can confidently identify. If the engine can't determine who published a page — what organization, what author, what their credentials are — citation confidence drops. Organization and Person schema give the engine a named entity with stable identifiers: @id, url, sameAs links to authoritative profiles (LinkedIn, Wikidata, Crunchbase).

Entity disambiguation is the cheapest lift on this list. It requires no content changes, no new pages, no ongoing maintenance. You add it once to your site's global schema layer and it applies to every page.

The mechanism: engines maintain entity graphs. When your Organization schema links to your Crunchbase profile and your LinkedIn page via sameAs, the engine can cross-reference those nodes and increase its confidence that citations from your domain are attributable to a real, identifiable entity. Unidentified domains get cited less, not because the content is worse, but because the engine is less willing to stake a citation on an ambiguous source.

For Person schema on author pages: include jobTitle, affiliation, and at least two sameAs links. An author with a verifiable identity gets cited more than an anonymous byline.

BreadcrumbList showed zero citation benefit across 340 pages

We ran BreadcrumbList schema across 340 pages for six months and measured citation rate before and after. The result was flat — no measurable lift in AI citation frequency.

The reason is structural. Citation engines don't use navigational hierarchy signals. They're not trying to understand where a page sits in a site's information architecture. They're trying to extract a specific answer to a specific query. BreadcrumbList provides no answer-extraction signal.

BreadcrumbList still earns sitelinks in traditional search results and contributes to crawl efficiency. Implement it for those reasons. Don't implement it expecting AI citation lift — the data says it won't come.

Frequently asked questions

What structured data type has the biggest impact on AI citations?

FAQPage schema has the biggest impact on AI citations. AI engines extract Q&A pairs verbatim into answer slots, making FAQPage markup the most directly consumable format for generative answer engines. Pages implementing FAQPage schema have shown a 2.3x lift in AI citation frequency compared to pages without it. No other schema type delivers that level of direct extraction into AI-generated responses.

Does structured data help with ChatGPT citations or only Google AI Overviews?

FAQPage and HowTo schema benefit Perplexity and Google AI Overviews most directly, since both systems actively parse structured markup to populate answer slots. ChatGPT's citation behavior is less dependent on schema markup — it relies more on content quality and entity recognition than on structured data signals. That said, Organization and Person schema still help with ChatGPT by resolving entity identity, making it easier for the model to attribute content to a credible, known source. Structured data is worth implementing for all three, but the ROI is highest for Perplexity and AI Overviews.

How often should you update your structured data?

Update your structured data whenever you update the underlying page content, and ensure the dateModified field reflects the actual change date. Pages updated within the past 90 days earned 56% more citations from Perplexity than stale pages, making freshness a measurable ranking signal for AI engines. Beyond reactive updates, audit your structured data at least quarterly to catch schema drift, deprecated markup, or new schema types that apply to your content. Keeping dateModified accurate is the single highest-leverage action in this process.