Schema Markup That AI Actually Cites: Beyond FAQ – Spice Up

I already wrote about FAQ schema in 2026 — short version: it’s still useful, just not the rich-snippet darling it once was. But that piece got a lot of follow-up questions like “okay, what about all the OTHER schema types?” Fair question.

Here’s the deal in 2026: Google quietly killed seven more schema types’ rich results in January, then dropped the March update that says schema must match the primary content topic of a page (no more sprinkling FAQ markup on product pages to game it). Meanwhile, AI engines like ChatGPT, Perplexity, and Claude are doing something different — they’re using your structured data to build entity graphs and decide who’s worth citing.

So which schemas still matter? Which ones AI actually reads? And which ones are dead weight you should rip out? Let’s get into it.

TL;DR

• Seven schema types lost rich results in January 2026: Course Info, Claim Review, Estimated Salary, Learning Video, Special Announcement, Vehicle Listing, Practice Problems
• HowTo is fully dead — desktop in 2023, mobile finished in 2024-2025. Stop using it
• March 2026 update: schema must match primary content topic. No more FAQ-on-product-pages tricks
• The schema types AI cares about most: Article/BlogPosting, Organization, Person, Product, BreadcrumbList, VideoObject, ImageObject, FAQPage, Event, Recipe
• The secret weapons: @id and sameAs properties build entity graphs that AI uses to decide citation confidence
• Honest reality check: Google says “no special schema needed” for AI Overviews. But studies show 71-82.5% of AI-cited pages use structured data. The correlation is real, even if causation is debated

The Honest Schema-and-AI Debate

Quick Answer: Google says structured data isn’t required for AI Overviews. But independent data shows 65-82.5% of AI-cited pages use schema markup, suggesting it correlates with — if not causes — citations.

Before I list a bunch of schema types, let’s address the elephant in the room. There’s a real disagreement happening in the SEO world right now, and you should know about it.

Google’s official line: “There’s no special schema.org structured data that you need to add” for AI Overviews or AI Mode. Plain and direct, straight from Google’s documentation.

Independent research: SE Ranking found 65% of pages cited by Google AI Mode and 71% cited by ChatGPT include structured data. Other analyses claim 82.5% of AI citations come from pages with structured data. Search Engine Land’s reality-check piece found no direct correlation when controlling for content authority — meaning schema-heavy sites weren’t cited more, but schema is also probably proxying for “well-built site with good content.”

My take? Schema isn’t a magic citation lever. But it’s:

1. Cheap to implement (free generators exist for everything)
2. Helps AI systems disambiguate entities — which does affect citations
3. Builds knowledge graph alignment, which AI uses for verification
4. Future-proofs your content as standards evolve

The cost of doing it right is low. The cost of skipping it could be missed citations. Just don’t expect schema alone to save bad content.

What Got Cut: The 2026 Schema Cleanup

Before adding new schema, audit what you have. Google has been aggressively pruning supported types.

Schema Type	Status in 2026	What to Do
HowTo	Fully deprecated (desktop 2023, mobile 2024)	Remove. Replace with Article or step-list in body content
Course Info	Removed January 2026	Replace with Article or Event (if scheduled)
Claim Review	Removed January 2026	Use Article with appropriate context
Estimated Salary	Removed January 2026	Use Article or JobPosting
Learning Video	Removed January 2026	Standard VideoObject still works
Special Announcement	Removed January 2026	Use Article or NewsArticle
Vehicle Listing	Removed January 2026	Use Product schema
Practice Problems	Removed January 2026	Use Article
FAQPage (off-topic)	Restricted March 2026 — must match primary topic	Only use on actual FAQ pages, not as bolt-on
Review (off-topic)	Restricted March 2026 — must match primary topic	Use only on dedicated review pages

The deprecated schemas don’t break anything — they’re just ignored. But carrying dead schema has a real cost: maintenance overhead, false sense of optimization, and confusion when something stops working. Just remove them.

The Schema Types That Still Earn AI Citations

Let’s get to what works. I’ll organize these by purpose, not just alphabetically — because how you use them matters more than what they’re called.

Article, BlogPosting, NewsArticle: The Content Foundation

Quick Answer: Article is the baseline for any text content AI might cite. BlogPosting is the right subtype for blogs, NewsArticle for journalism. All three help AI systems attribute quotes correctly.

This is the most important schema for content publishers, full stop. Article schema (and its subtypes) tells AI systems: this is original written content, here’s who wrote it, here’s when it was published, here’s what it’s about.

The connection to E-E-A-T is direct. Article schema forces you to declare an author, a publish date, and an organization. Those are exactly the signals Google uses for Experience-Expertise-Authoritativeness-Trustworthiness — and they’re the same signals AI engines use to decide whether to cite you.

Here’s a minimum-viable Article schema with author and publisher properly linked:

Notice the @id values — that’s how you tell AI “this Person, this Organization, this Article are stable entities I’m referencing across my whole site.” More on that in a minute.

Organization Schema: Your Entity Anchor

Quick Answer: Organization schema is the foundation of your entity graph. AI engines like Perplexity and ChatGPT use it to verify your brand and decide whether to cite you with attribution.

If you only do one piece of site-wide schema, do this one. Organization schema lives once (typically on your homepage or as a sitewide blob) and gets referenced everywhere via @id.

The critical fields:

• name — your canonical brand name
• url — your homepage
• logo — full URL to a logo image
• sameAs — array of authoritative profile URLs (Wikipedia, Wikidata, LinkedIn, GitHub, X, etc.)
• description — one-sentence brand summary
• foundingDate, founder — entity context

The sameAs trap: only include profiles you actively maintain. An abandoned 2014 Tumblr account weakens your entity signal because AI systems will fetch it and find nothing recent. If you haven’t posted in six months, leave it out.

Person Schema: Author Authority for E-E-A-T

Quick Answer: Person schema is how you tell AI “this writer is a real human with verifiable expertise.” It’s becoming critical as AI engines weigh author credibility for citations.

Author Schema (which uses Person) is having a moment. AI engines need clear attribution to credit original sources, and Person schema is the mechanism that links a writer to their credentials, their other published work, and their external presence.

Like Organization, define each Person once and reference via @id from every Article they wrote. Critical fields:

• name — full name
• url — author bio page on your site
• jobTitle — role
• worksFor — link to Organization via @id
• sameAs — external profiles (LinkedIn, GitHub, Mastodon, ORCID, Wikipedia)
• knowsAbout — array of expertise areas (helps with topic matching)
• alumniOf — credentials angle

The payoff: when AI engines cite content, they increasingly mention authors by name. Person schema with proper @id referencing means your name accumulates citation authority across every piece you publish, instead of being treated as five different people on five different posts.

Product Schema: Critical for AI Shopping

Quick Answer: Product schema is the highest-ROI schema for e-commerce in 2026 — both for traditional rich results and for citations in ChatGPT shopping, Perplexity Shopping, and Amazon Rufus.

If you sell anything, this is non-negotiable. AI shopping exploded in 2025-2026 with OpenAI’s Agentic Commerce Protocol (powering ChatGPT Instant Checkout via Stripe), Google’s Universal Commerce Protocol (with Walmart, Target, Shopify, Etsy), and Perplexity Shopping all citing products from structured data.

The catch in 2026: AI engines cite brands and model names, not URLs. So Product schema isn’t just about rich results — it’s about being identifiable as a citeable product entity.

Required fields for AI visibility:

Don’t skip gtin13, sku, or brand — these are the fields that let AI systems disambiguate your product from competitors with similar names.

BreadcrumbList: The Underrated One

Quick Answer: BreadcrumbList tells AI exactly where a page sits in your site hierarchy. Critical for context and topical relevance, often overlooked.

Boring. Easy. Effective. BreadcrumbList doesn’t get rich results in the visual sense, but it gives AI engines a map of your site’s information architecture. When ChatGPT cites a page on “self-hosted RAG,” knowing it sits in portalZINE → AI → Self-Hosting → Self-Hosted RAG matters for topic disambiguation.

Most CMSs auto-generate this. If yours doesn’t, the manual implementation is trivial. Just do it.

VideoObject and ImageObject: Multimodal AI Wins

Quick Answer: VideoObject and ImageObject schemas help your media surface in AI multimodal answers, image carousels, and video citations across Google AI Overviews and Perplexity.

AI search is increasingly multimodal. ChatGPT shows images, Google AI Overviews embed videos, Perplexity references screenshots. Without VideoObject and ImageObject markup, your media is just decoration as far as AI is concerned.

For video — especially if you self-host or aren’t on YouTube:

For ImageObject, the high-leverage fields are contentUrl, creator, license, and caption. License especially matters — AI engines that respect rights are more likely to cite images with explicit licensing.

FAQPage: Still Useful, More Constrained

Quick Answer: FAQPage schema still drives AI citations and is featured in Answer Engines, but Google’s March 2026 update means it only works on actual FAQ pages — not bolted onto product or service pages.

I covered this in detail in my FAQ schema 2026 piece, but the headline updates:

• Still highly cited by AI: FAQPage remains one of the highest-leverage schemas for ChatGPT, Perplexity, and Google AI Overviews
• Google rich results restricted to government/health authority sites since 2023
• March 2026 update says FAQ schema only counts when the page’s primary topic is the Q&A — no more sprinkling it on every blog post for SEO juice
• Optimal answer length: 40-60 words per answer for AI extraction (long enough to be substantive, short enough to be quotable)

Specialty Schemas: Event, Recipe, JobPosting

Use these only when they actually fit your content. Don’t shoehorn.

• Event — Conferences, webinars, scheduled sessions. AI engines surface dates, locations, ticket links
• Recipe — Food content. Still gets rich snippets with ratings, cook times, ingredients. AI cooking assistants pull from this directly
• JobPosting — Job listings. Still alive after the EstimatedSalary cull. Required for Google for Jobs visibility
• Review — Use only on actual review pages. The March 2026 update killed off-topic Review markup

The Real Power-Ups: @id and sameAs

Quick Answer: @id gives entities stable identifiers across your site so AI systems recognize they’re the same thing. sameAs links your entities to authoritative external sources for verification.

This is the part most schema guides skip and it’s the most important. @id and sameAs are how you go from “scattered schema blobs” to “an entity graph that AI engines actually use.”

The @id Trick

Without @id, your CEO mentioned across five articles appears to AI as five different people named the same thing. Authority doesn’t accumulate. With consistent @id:

Now AI engines understand: same person, every reference. Citation authority compounds across all 200 articles you’ve written.

The sameAs Connection

The sameAs property is the bridge between your entity graph and the broader web. Link your Organization to its Wikidata entry, your Person to LinkedIn and ORCID, your Product to manufacturer pages. AI engines use these to verify entity claims.

Best targets for sameAs (in rough order of impact):

1. Wikidata — the structured Wikipedia. Highest authority for AI entity verification
2. Wikipedia — if you have an entry, link it
3. LinkedIn — strong for Person and Organization
4. GitHub / GitLab — strong for developer/tech entities
5. Crunchbase — strong for company entities
6. ORCID — strong for academic Person entities
7. Official social profiles you actively maintain

Get yourself a Wikidata entry if you don’t have one. It’s free, it’s slow to approve, and it pays dividends for years.

Phased Implementation Plan

You probably can’t do all of this at once. Here’s the priority order I’d actually recommend.

Phase	Action	Priority
Phase 1	Audit existing schema. Remove deprecated types (HowTo, the 7 January casualties)	HIGH — Do first
Phase 1	Site-wide Organization schema with sameAs links	HIGH — Foundation for everything else
Phase 1	Person schema for every author with @id reference to Organization	HIGH — E-E-A-T foundation
Phase 1	Article/BlogPosting on every content page with author + publisher @id refs	HIGH — Content baseline
Phase 2	BreadcrumbList everywhere	MEDIUM — Easy win, often auto-generated
Phase 2	Product schema with brand, GTIN, offers (e-commerce only)	MEDIUM-HIGH for e-com
Phase 2	FAQPage on actual FAQ pages only (not bolted on)	MEDIUM
Phase 3	VideoObject and ImageObject for media-heavy pages	LOW-MEDIUM
Phase 3	Specialty schemas (Event, Recipe, JobPosting) where applicable	LOW — Only if relevant
Phase 3	Get Wikidata entry for your brand and key authors	LOW — Long-term entity authority

Free Tools to Validate and Generate

• Schema.org Validator — The official validator. Use this first
• Google Rich Results Test — Tells you whether your schema is eligible for rich results
• Google Search Central Docs — Official guidance per schema type
• Schema.dev — Free generator with copy-paste JSON-LD
• Hyperleap Schema Generator — Free generator with broad type support
• iLoveSchema — Free article and blog post schema generator

If you’re on WordPress, the popular SEO plugins (Yoast, Rank Math, All in One SEO) all include schema generators in their free tiers. They handle the connecting @id references automatically, which is the part most people get wrong manually.

What Doesn’t Work (Save Yourself the Time)

• Sprinkling FAQ schema on every page — March 2026 update killed this. Schema must match the page’s primary topic now
• Using HowTo schema — fully dead, just remove it
• Including dead social profiles in sameAs — weakens entity signal. Only active, maintained accounts
• Skipping @id values — biggest schema sin in 2026. Without them, your entities don’t accumulate authority
• Adding schema to game rankings — Google explicitly states schema isn’t a ranking factor. Use it for AI citation eligibility, not ranking tricks
• Inventing your own custom types — stick to the schema.org vocabulary. AI engines were trained on canonical types

The Bottom Line

Schema markup in 2026 isn’t dead — it’s just not what it was in 2020. The visual rich-snippet era is mostly over for non-authority sites. The new game is entity graphs, AI citations, and structured signals that help machines understand who you are, what you publish, and why you’re worth citing.

The schema types that survive the cleanup and earn AI citations are the ones that describe core entities (Organization, Person, Article, Product) and connect them via @id and sameAs. Everything else is either gone (HowTo, the 7 January casualties), constrained (FAQPage, Review), or specialty (Event, Recipe).

If you’ve been ignoring schema because “Google says it’s not a ranking factor” — that’s true, but increasingly beside the point. AI engines cite content, and they cite content from sites whose entities they can verify. Schema is how you become verifiable.

Start with Organization, Person, and Article using proper @id references. Add sameAs links to authoritative profiles. Validate with the Schema.org Validator. Audit deprecated types and rip them out. That’s the 2026 schema playbook.

FAQ