JSON-LD and Google AI Search: What Really Matters

No: JSON-LD does not decide whether Google cites your page in AI-powered search features. It can help Google interpret eligible structured information, but a citation depends on whether Google finds, indexes, understands, and chooses to use the page for a particular query; adding markup cannot guarantee visibility or citation.

What JSON-LD does—and does not do

JSON-LD is a format for expressing structured data in a page, usually inside a script element. It can describe entities and relationships—for example, that a page is a recipe, an organization, an event, or a product. Search engines may use valid, relevant structured data to understand content and, when applicable, show an enhanced search result.

That is different from deciding which sources appear in an AI-generated answer. Structured data is not a command to cite a URL, a ranking shortcut, or a substitute for the page itself. A page can have valid JSON-LD and never be cited; a useful, indexable page can be cited without JSON-LD.

Google’s public guidance has consistently treated structured data as a way to make content eligible for supported search features, not as a guarantee that a feature will appear. Eligibility depends on the relevant feature’s requirements, and Google ultimately controls whether and how it displays results.

How Google may select sources for AI features

Google does not publish a complete, fixed formula for selecting sources in every AI-generated search experience. In broad terms, its search systems need to discover pages, crawl them where permitted, index and interpret their content, and determine whether the information is useful for the user’s query. AI features can draw on search systems and may show links to supporting pages, but their presentation and availability can vary.

For a page to have a realistic chance of being considered, it generally needs to be accessible to Google and contain information that directly helps answer the query. Technical eligibility is only a starting point: a page that is blocked from crawling, marked not to be indexed, inaccessible, or too thin to be useful is unlikely to serve as a dependable source.

Selection is query-specific. A detailed, trustworthy explanation may be relevant to one question and irrelevant to another. Even when Google indexes a page, it does not follow that the page will appear for every related query, in an AI answer, or with a citation.

Where structured data fits

JSON-LD can clarify facts that are already visible on the page. For instance, a cooking page might use appropriate recipe structured data to identify its name, ingredients, and preparation details. A business page might identify the organization behind the content. The markup should accurately reflect the page and follow Google’s documentation for the particular structured-data feature.

It is best viewed as supporting context rather than a citation mechanism. Structured data may reduce ambiguity about what a page contains, but it does not make weak or unsupported claims reliable. Nor should it be used to put information in markup that readers cannot find in the visible content. Inaccurate, misleading, or irrelevant markup can make a page ineligible for a rich result and may undermine trust in the data.

Use JSON-LD when it genuinely describes the content and there is a relevant supported feature or a clear reason to provide structured information. Do not add a schema type just because it sounds likely to trigger an AI citation.

What to prioritize for a better chance of being cited

There is no checklist that guarantees citations, but practical search fundamentals are more meaningful than adding markup alone. Make the page easy to access, understand, and evaluate, and answer the reader’s question with original, specific information.

Check crawlability and indexing: Make sure the URL is not unintentionally blocked by robots directives, an access restriction, or a noindex instruction. Use Search Console to inspect the page and review any reported indexing issues.
Answer the query directly: Put the central answer where readers can find it, then explain the evidence, qualifications, and relevant details. Avoid burying the useful material beneath generic introductory text.
Show who is responsible: Provide clear authorship or organizational information where appropriate, and make it possible to assess the source and its editorial standards. These practices support reader confidence; they do not guarantee a citation.
Keep claims grounded: Cite primary or otherwise reliable sources for factual claims, distinguish established information from uncertainty, and update material when it becomes outdated.
Improve the page experience: Ensure important content is available to crawlers and readers, works on mobile devices, and is not obscured by intrusive elements or broken page components.
Use structured data accurately: Validate the markup, keep it consistent with visible content, and check that it meets the documentation for the feature you are targeting.

These are sound publishing and technical practices, not a secret recipe for inclusion. Search systems can change, and the best source for a query may differ as the query, available pages, and search experience change.

Common misconceptions and troubleshooting

“I added Article schema, so Google should cite the page.” Article markup may help describe an article, but it does not instruct Google to display a citation. If the page is not appearing, check whether it is indexed and whether it offers a distinctive, relevant answer before changing schema.

“A rich result means an AI citation is next.” Rich results and AI citations are different outcomes. Eligibility for one search feature does not establish eligibility or selection for another.

“A successful validation test proves Google will use it.” A test can identify syntax problems or assess eligibility for supported features. It cannot promise that Google will show a result, index the URL, or cite it in an AI answer. A valid JSON-LD block can still describe the wrong page or conflict with its visible content.

If a page is missing from search, begin with indexing diagnostics rather than repeatedly rewriting markup. Confirm that Google can fetch the URL, inspect the indexing status, review canonicalization and noindex settings, and check whether the page provides substantial value beyond material already available elsewhere. After fixing a problem, recrawling and search updates may take time; a request for indexing is not a guarantee of inclusion.

How to measure results realistically

Search Console can help assess ordinary Google Search visibility, including impressions, clicks, and indexing information. Its reports do not necessarily provide a complete, separate accounting of every AI-generated citation or every appearance of a page in an evolving search interface. Do not infer citation success simply from an increase in traffic, or failure from the absence of a clearly labeled report.

For a practical review, record the pages and queries that matter, inspect search results periodically, and look for changes in indexed status and organic performance. Treat spot checks as observations, not definitive measurements: results can vary by query, location, time, and interface. Keep the content useful regardless of whether a particular AI feature sends visitors.

Frequently Asked Questions

Does Google require JSON-LD to cite a website in AI search?
No. JSON-LD is not a general requirement for a page to be considered as a source. A page’s accessibility, indexing, relevance, and usefulness matter; supported structured data may help describe eligible content but does not guarantee a citation.

Can schema markup increase the chance of an AI citation?
It may help Google interpret certain information when the markup is accurate and relevant, but Google does not promise that schema will increase the chance of an AI citation. There is no markup type that forces a citation.

Can a page appear in an AI answer without structured data?
Yes. Structured data is not a prerequisite for every search appearance or citation. Google can interpret a page from its accessible content and other search signals.

Which schema type should I use for AI Overviews?
Use the structured-data type that accurately describes the page and is supported for the relevant search feature. Do not select a type solely to target AI Overviews; no schema type guarantees inclusion there.

How can I tell whether Google indexed my page?
Use the URL inspection tools in Google Search Console to review the page’s reported indexing status and any detected issues. Being indexed means the page can be eligible to appear in search; it does not promise a ranking or citation.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.

More cards in this drawer

The whole Numbers and Powers drawer →