In short: Schema markup describes visible content in a machine-readable form, usually JSON-LD. Choose a type that genuinely matches the page, keep markup and copy aligned, then inspect the rendered document. Accurate markup can reduce ambiguity for crawlers, but it does not guarantee a ranking or an AI citation.
This practical guide explains how to add structured data without marking up facts that are absent from the page or claiming a direct effect that has not been demonstrated.
For the differences between models, engines and practices, see the definitions in our glossary: LLMs, AI answer engines, ChatGPT, Gemini, Perplexity, Answer Engine Optimization, Claude.
What is schema markup for?
Structured data gives an explicit description of elements already visible on a page: an article, organization, offer, breadcrumb, question and answer, or ordered steps. It does not replace the page, fix inaccurate information or compensate for thin content.
The exact way each AI product uses structured data is not publicly documented consistently. It is therefore not sound to claim that a schema.org type causes a citation. The safe rule is simpler: make facts visible, accessible and consistent.
Choose the type that matches the page
| Real page purpose | Possible type | What must be visible |
|---|---|---|
| Editorial guide or analysis | Article | Headline, body, publisher or author, real dates |
| Organization page | Organization | Name, official site and verifiable public details |
| Questions and answers | FAQPage | Every question and its full answer |
| Repeatable process | HowTo | Ordered steps a reader can follow |
| Navigation | BreadcrumbList | The same visible breadcrumb |
Start with the page’s job, not a list of types to stack. Adding FAQPage without a visible FAQ, or Product to an advice article, creates a mismatch.
FAQPage: describe a visible FAQ
FAQPage fits when questions and answers genuinely appear on the page. According to the official Google Search Central changelog entry dated May 8, 2026, FAQ rich results no longer appear in Google Search as of May 7, 2026. Do not choose this markup in expectation of a search-result enhancement. Use it only to describe a visible, accurate FAQ.
Implementing FAQPage
- Write the complete, verifiable visible answer first.
- Mirror that exact question and answer in JSON-LD.
- Remove or update the markup when the visible FAQ changes.
- The number of questions is an editorial choice: include only those that serve the page.
HowTo, Organization and Article: practical use
Implementing HowTo
Use HowTo when a reader can complete a task through a full, ordered sequence. Show the same steps in the body, including conditions that change the outcome. A list of tips is not automatically a process.
Implementing Organization
State the official name, official URL and only external profiles that represent the same entity. The sameAs property declares a relationship; it does not prove that an AI system will always join those pages or prevent every ambiguity.
Implementing Article
For editorial content, mirror the headline, publisher or author, and dates actually shown to readers. Change dateModified after a material update, not to simulate freshness.
A minimal JSON-LD example
This educational example describes a fictional organization. Replace every value with a fact that is genuinely shown on the page: {"@context":"https://schema.org","@type":"Organization","name":"Example Company","url":"https://example.com"}.
Your template should place this object in a JSON-LD script element. The pattern is reusable; the values are only valid when they match visible content.
Validate markup before publishing
Use Schema Markup Validator to check syntax and vocabulary. Use Google’s Rich Results Test only when you target a supported Google feature.
Inspect the rendered page too. A static HTML fetch can miss JSON-LD injected by JavaScript. If the block is absent from the rendered document, crawlers that inspect that document do not receive it that way.
Measure without jumping to conclusions
Before a change, record the URL, stable queries, metrics consulted and collection date. After publication, compare the same page and queries over a window chosen in advance. Recrawling and indexing vary by engine, accessibility and crawl frequency: a fixed delay does not prove an effect in generated answers.
The most useful check remains consistency: visible copy, HTML structure, structured data and public identity should state the same facts.
Two documented reference points for this guide
The work queue created on November 1, 2026 records 515 impressions and 0 clicks for this guide’s French URL in the consulted Google Search Console data. The English version of the same resource records 13,865 impressions and 1 click. These are observations for those URLs and that collection, not proof that a schema type caused them.
Our public barometer offers a second reference point: from September 2025 to June 2026, 104 audits across 65 brands analysed 6,820 answers from 44 models. In a 13-brand sub-sample, 106 of 657 answers, or 16%, mentioned the audited brand without naming it in the question. That sub-sample does not represent every brand; it shows why clean markup cannot replace verifiable facts or useful content.
Editorial checks before publishing
- Confirm that every JSON-LD property matches a visible, current element.
- Choose one primary type for the page’s real job, then add only justified types.
- Check rendered HTML, not only a template or source file.
- Keep the URL, collection date and compared queries before drawing conclusions.
Read our complete AEO audit guide for a related measurement workflow.
Frequently asked questions
What is schema markup for AI?
Schema markup is structured data, usually JSON-LD, that describes a page’s visible content in a machine-readable form. It can identify an article, organization, FAQ or process, but it does not replace the content itself.
Which schema types should a page use?
Match the type to the page’s real purpose: Article for editorial content, Organization for an entity, FAQPage for visible questions and answers, and HowTo for a complete ordered process. Adding a type that does not match the content creates a mismatch, not an advantage.
How do you implement schema markup?
Add JSON-LD that mirrors visible facts, validate the syntax with Schema Markup Validator, and inspect the rendered document with Google’s Rich Results Test when targeting a Google feature. A static HTML fetch can miss markup injected by JavaScript.
Does schema markup improve AI visibility?
There is no reliable basis for claiming that schema directly improves ranking or guarantees an AI citation. Accurate markup can reduce ambiguity when it describes useful, accessible and visible content. It clarifies content; it does not substitute for it.
What mistakes should you avoid with schema markup?
Avoid marking up prices, ratings, dates, authors or answers that visitors cannot see; leaving facts stale; choosing a type unrelated to the page; and skipping rendered-page validation after publication.
Questions and answers
What is schema markup for AI?
Schema markup is structured data (usually JSON-LD) that describes your visible page in a machine-readable form: article, organization, product, Q&A or process steps. It helps search and AI systems interpret what is already on the page; it does not replace the content itself.
Which schemas matter most for AEO?
Match the schema to the page's real job: Article for editorial content, Organization for your entity, FAQPage for genuine questions and answers, HowTo for a complete visible process. Adding types that don't match visible content creates a mismatch, not an advantage.
How do you implement schema markup?
Add JSON-LD that mirrors the visible page, validate syntax with Schema Markup Validator, and test the rendered document with Google's Rich Results Test when you target a Google feature. A static HTML fetch can miss JavaScript-injected markup, so always check the rendered page.
Does schema markup really improve AI visibility?
There is no reliable basis for claiming schema directly improves ranking or guarantees an AI citation. Accurate markup reduces ambiguity for crawlers and answer engines when it describes useful, accessible content. Fix the content first; schema clarifies, it doesn't substitute.
What mistakes should you avoid with schema markup?
Marking up claims visitors can't see (ratings, prices, answers), leaving dates or facts stale, mismatching the type to the page's purpose, and skipping rendered-page validation after publishing.
Agencies, consultants and in-house teams: compare the answers generated for your own scenarios, languages and selected models. Our infrastructure, your scenarios, your expertise.