How to use schema markup to get cited by AI (2026)
Use schema markup to help AI answer engines cite you: pick the right JSON-LD types, verify your entity, keep markup matched to visible content, and validate it.
Schema markup — the schema.org vocabulary you add to a page as JSON-LD — is a labeled, machine-readable copy of what the page already says: this is the organization, this is the author, this is the product, these are the facts. Answer engines like ChatGPT, Perplexity, Google's AI Overviews, and Gemini read that label to parse your content faster, disambiguate your entity, and cross-check a claim before they lean on it. Used right, it makes your page easier to extract and your identity easier to verify.
Be clear about what it does and does not do, because the internet oversells it. Schema does not directly cause a citation and it will not rescue a thin or unreachable page — controlled tests have found little standalone lift on pages that were already widely cited. What it does is remove ambiguity: it gives the retriever an unambiguous version of your facts and a way to tie your content to a credible, consistent entity. As Search Engine Journal put it, schema doesn't create trust, it makes trust verifiable. Treat it as one necessary layer under useful writing, real authority, and third-party corroboration — not a switch that flips citations on. If a crawler cannot reach the page at all, fix that in [make content visible to AI search](/how-to/make-content-visible-to-ai-search) before any of this matters.
The steps
Map the page to the schema types that actually fit it. Match the markup to what the page is, never to what you wish it ranked for. The high-value types for AI parsing are Organization (who publishes this), Article or BlogPosting (the content itself), Person (a named author), FAQPage or QAPage (a real question set), and Product, Service, or LocalBusiness for commercial and local pages. A HowTo procedure can carry HowTo markup. One page usually deserves two or three of these nested together, not a pile of every type you can find.
Write it as JSON-LD in the page head. Use JSON-LD — a single <script type="application/ld+json"> block — rather than inline microdata or RDFa. It is the format Google recommends and the cleanest for a machine to lift whole, and it keeps the markup separate from the visible HTML so it is easy to audit. Most CMS and SEO plugins emit it for you; if you hand-write it, keep one block per entity and reference them together by @id.
Nail the entity: Organization and Person, with sameAs. The single most useful thing schema does for AI search is pin down who you are. Fill Organization with your exact legal name, logo, and description, and give it a sameAs array linking to the profiles that corroborate you — LinkedIn, Wikipedia or Wikidata, Crunchbase, your verified social accounts. Do the same for the author with Person markup and a real, linked bio. An engine that sees the same entity described consistently across the open web treats it as an identity it can trust and name.
Make every schema value match the visible page. Never let the markup claim something a human cannot see on the page. If your JSON-LD lists a price, a rating, an author, or a publish date, that exact value has to appear in the visible content too. Mismatches — schema saying one thing, the page another — are the fastest way to get your structured data ignored or flagged, and they read as manipulation to the same systems you are trying to earn trust from.
Keep facts consistent across every surface, not just the page. Schema is one of four surfaces that have to agree: the visible page, the markup, your platform of record (Google Business Profile, a Merchant feed, your LinkedIn), and third-party sources that mention you. An engine gains confidence when your name, category, address, and one-line positioning are identical everywhere and hesitates when they conflict — "Suite" on one, "Ste" on another is enough friction. Standardize the entity facts once and mirror them across all four.
Don't expect a SERP rich result from FAQ or HowTo markup. Google narrowed FAQ rich results to a small set of authoritative health and government sites in 2023 and retired HowTo rich results the same year, then dropped FAQ rich results from Search entirely on May 7, 2026 — no site gets that SERP feature anymore. The markup is still valid schema.org and still helps machines parse your question-and-answer structure — AI systems in particular use those Q&A pairs — so keep it if it matches real content, but add it to aid extraction, not to win a search feature that no longer exists.
Validate before you ship. Run every page through the Schema Markup Validator (schema.org) and Google's Rich Results Test. Fix errors and warnings, confirm the parsed output matches your intent, and check that the types you added are the ones that rendered. Broken or malformed JSON-LD is worse than none — a parser that chokes on your block gets nothing, so a passing validation is the floor, not the finish line.
Pair schema with the content and authority it points at. Structured data is a signpost; it needs something real to point to. After the markup is clean, do the work that makes the page worth citing — a standalone answer up top, cited statistics and quotations, a credible named author, third-party mentions — covered in [optimize a page to get cited by AI search](/how-to/optimize-a-page-to-get-cited-by-ai-search). Then re-test your target prompts in the engines over the following weeks and keep both the content and the entity facts current, because schema that labels a stale page just labels it accurately.
Common gotchas
Treating schema as a citation switch. It removes ambiguity and verifies your entity; it does not manufacture authority or rescue a thin, unreachable, or unhelpful page. Controlled research found little standalone lift on already-cited pages — it is necessary, not sufficient.
Marking up claims the page does not show. If your JSON-LD asserts a price, rating, author, or date that is not visible on the page, engines discount the markup and may flag it. Every schema value must have a visible counterpart.
Skipping sameAs. Organization and Person markup without a sameAs array to corroborating profiles wastes the one thing schema does best — pinning your entity to a verifiable identity across the open web.
Adding FAQ or HowTo markup expecting a rich result. Google retired HowTo rich results and restricted FAQ rich results to authoritative health and government sites in 2023, then dropped FAQ rich results for every site on May 7, 2026. Keep the markup only where it matches real content and only for machine parsing, not a SERP feature.
Never validating. Malformed JSON-LD parses to nothing. Run the Schema Markup Validator and Rich Results Test after every change, and re-check when a CMS or plugin update might have altered the output.
Inconsistent entity facts. Schema that says one name or address while your Google Business Profile, LinkedIn, and citations say another creates exactly the ambiguity you added markup to remove.
Where Kompozy fits
[Blog Articles](/how-to/write-content-that-performs-in-ai-search) in Kompozy already carry their own Article and FAQPage JSON-LD, generated automatically at render time — for WordPress and Custom Webhook destinations that markup ships as-is; on GHL Blog it is stripped before publish because GHL's own template emits its own Article schema and two copies would just create duplicate-content noise. So the page-level piece this tutorial walks through is partly handled already. What schema on one page can't do — and where Kompozy matters more — is the other half of the citation problem: an answer engine confirms an entity by seeing it described the same way everywhere it looks, not just in one markup block. That cross-surface consistency is what an Organization or Person sameAs array is trying to assert, and it's what Kompozy produces at scale. The [Persona Brief](/glossary/persona-brief) fixes your exact name, one-line positioning, and brand facts once, and every output inherits them: Blog Articles plus Text Posts, Carousel Posts, Infographic Photos, and [Persona Shorts](/glossary/persona-shorts) whose named on-camera author reinforces the same Person entity your markup points at. Publish those with a real byline and consistent bio and you feed the sameAs graph the corroborating mentions that turn schema's claim into something an engine can verify. [Autopilot](/glossary/autopilot) schedules the set across the eight social platforms plus blog and email through a per-post review gate, so the entity facts stay identical on every surface instead of drifting the moment a dozen posts are written by hand — the ambiguity schema exists to remove, closed on the platforms schema never reaches. The framework tying markup, content, and entity together is in the guide on [AI search citation optimization](/guides/ai-search-citation-optimization). Starter ($99/mo, 5,500 credits) fits a solo creator hardening one brand's footprint; Pro ($299/mo, 18,000 credits) suits a business keeping its entity consistent across every channel; Enterprise is custom for agencies.
Frequently asked questions
Does schema markup get you cited by AI?
Not directly. Schema markup does not cause a citation; it makes your facts machine-readable and your entity verifiable, which helps answer engines parse and trust your page. Controlled tests found little standalone lift on pages that were already widely cited, so treat it as one necessary layer beneath useful content, real authority, and third-party corroboration — it makes you eligible and easy to extract, it does not make the model choose you.
Which schema types matter most for AI citations?
For AI parsing the highest-value types are Organization, Article or BlogPosting, Person for a named author, FAQPage or QAPage for a genuine question set, and Product, Service, or LocalBusiness for commercial and local pages — all in JSON-LD. Most pages deserve two or three nested together that match the content, not every type available. Organization and Person with a sameAs array do the most work because they pin down who you are.
Should I still use FAQ schema in 2026?
Only where it matches real question-and-answer content, and not for a search feature. Google restricted FAQ rich results to authoritative health and government sites and retired HowTo rich results in 2023, then dropped FAQ rich results for every remaining site on May 7, 2026, so no site gets a SERP feature from either anymore. The markup remains valid schema.org and still helps machines — including AI systems — parse your Q&A structure, so keep it for extraction, not for a rich result.
What is JSON-LD and why is it recommended?
JSON-LD is a way to write schema.org markup as a self-contained JavaScript block in the page head, separate from the visible HTML. Google recommends it, and it is the cleanest format for a machine to lift whole and for you to audit. The alternatives, inline microdata and RDFa, tangle the markup into the page body and are harder to validate, so JSON-LD is the default for both SEO and AI-search work.
How do I verify my schema is working?
Run the page through the Schema Markup Validator on schema.org and Google's Rich Results Test, fix every error and warning, and confirm the parsed types match what you intended. Malformed JSON-LD parses to nothing, so a clean validation is the baseline. Beyond that, test your target questions in ChatGPT, Perplexity, Gemini, and Google AI Overviews over the following weeks to see whether extraction and entity recognition actually improved.