Schema markup is back in a lot of AI search proposals, often with a promise the evidence does not support: add a few JSON-LD blocks and AI systems will cite the site more often.
Google’s guidance is much narrower. Its AI features documentation says there is no special schema or other machine-readable file required for AI Overviews or AI Mode. The same technical and content requirements used for regular Search still apply.
That does not make schema useless. Structured data still has documented jobs in Search. It can help Google understand specific details on a page and can make eligible pages available for supported search features. The useful question is not “Which schema ranks in AI?” It is “Which markup accurately describes this page and has a documented use?”
TL;DR
- Google does not require special structured data for AI Overviews or AI Mode.
- Use schema that matches the page and a documented search feature. There is no universal “AI schema stack.”
- Organization and Article cover common business and editorial details. Product, LocalBusiness, Event, VideoObject, JobPosting, and other types matter when the page and business qualify.
- Person can identify an author, but it is not a documented E-E-A-T score or AI citation switch.
- Google retired FAQ rich results on May 7, 2026. Do not present FAQPage as a special AI citation signal.
What schema can and cannot do for AI search
Google documents structured data as a way to provide explicit clues about a page and to become eligible for certain rich results. Eligibility is the key word. Valid markup does not guarantee that Google will show a feature, and it does not replace the ordinary requirements for crawling, indexing, quality, and policy compliance.
For AI features, the documented limit is just as important. Schema cannot force an AI Overview, improve a citation by itself, or tell ChatGPT what to say about a business. Other providers do not reveal enough about citation selection to turn “the markup is available in the HTML” into a proven ranking claim.
Schema may still be worth testing as part of a broader AI visibility program, but any extraction or citation effect should be labeled as a hypothesis and measured separately from its documented Search uses.
Choose schema by the page’s job
There is no responsible priority list that works for every site. A publisher, local clinic, ecommerce store, and software company have different pages and different eligible features.
| Page or goal | Markup to consider | Documented use | Important limit |
|---|---|---|---|
| Home or about page | Organization | Helps Google understand administrative details such as the organization’s name, URL, logo, and identifiers | It does not create authority or guarantee a knowledge panel |
| News or blog article | Article, BlogPosting, or NewsArticle | Supplies headline, image, author, and publication date details for article features | The content and author must be visible and accurately represented |
| Product page | Product plus eligible offer or review properties | Can qualify a page for product rich results or merchant listing experiences | Follow the separate rules for reviews, prices, availability, and variants |
| Physical business location | A suitable LocalBusiness subtype | Supplies business details for eligible local search experiences | It does not replace a claimed and accurate Google Business Profile |
| Event, video, recipe, job, or course-list page | The matching supported type | Can make an eligible page available for that feature | Required properties and eligibility rules differ by feature; Google’s course feature applies to lists of courses from one provider |
| Site navigation | BreadcrumbList | Can help Google display breadcrumb context in search results | It should represent the path users can understand on the site |
Google’s Search Gallery is the practical place to start. If a page type has documented properties and a feature your audience could actually see, implement it. If it does not, do not invent an AI benefit to justify the work.
Where Person and FAQPage fit
Person is useful when it truthfully identifies an author or the subject of a profile page. In Article markup, Google allows a Person or Organization as the author and recommends linking to a page that identifies that author. That is a clean authorship implementation. It is not evidence of an E-E-A-T score, a ranking boost, or preferential AI citation.
Google retired FAQ rich results on May 7, 2026 and removed the feature documentation in June. This supersedes the older restriction to government and health sites. FAQPage remains valid Schema.org vocabulary, but it no longer produces that Google feature. Keep useful visible FAQs and do not promise an AI citation benefit from the markup.
The one rule that keeps schema safe
Schema describes what is on the page. Never what you wish were on it.
Reviews that do not exist, hidden FAQs, an expired offer marked as available, or a dateModified that changes on every deploy can violate Google’s structured data guidelines or make the page ineligible for a feature. Passing a syntax test does not make a misleading claim acceptable.
Use dateModified only when the main content received a meaningful update. Google may use several page signals when it estimates a publication or update date. No official source says dateModified is a universal AI citation signal.
Implementation: the practical path
- Map templates to eligible types. List the actual templates on the site, then match each one to Google’s supported features. Do not start from a generic GEO checklist.
- Prefer JSON-LD for new work. Google supports JSON-LD, Microdata, and RDFa, but generally recommends JSON-LD because it is easier to maintain. There is no AI reason to rebuild valid Microdata.
- Generate markup from the same data as the page. Product price, author, availability, and dates should not be copied into a separate field that quietly goes stale. My schema markup generator prompt can help draft a block, but the page remains the source of truth.
- Validate syntax and eligibility. Use Google’s Rich Results Test for supported Google features and Schema.org’s validator for vocabulary. Then compare every property with the visible page.
- Check the rendered HTML. Make sure the production page contains the markup and that scripts, consent tools, or templates did not strip it.
- Measure the documented outcome. Watch the relevant Search Console enhancement report and actual search appearance. If you also test AI citations, record that as an experiment and do not attribute a change to schema without a controlled comparison.
The takeaway
Schema still deserves a place in technical SEO, but its value is specific. It describes eligible content, supports documented Search features, and gives teams a consistent way to maintain facts in code.
For Google’s AI features, there is no special markup requirement. For ChatGPT and other answer engines, citation effects remain unproven unless the provider documents them or your own test isolates them. Implement the schema the page qualifies for, keep it accurate, and spend the rest of the effort on content people can verify and use.
Frequently asked questions
Does schema markup help with AI search visibility?
Google says no special schema or other machine-readable file is required for AI Overviews or AI Mode. Standard structured data can still help Google understand eligible page details and enable supported search features, but Google does not document it as an AI citation or ranking lever. Use it because it accurately describes the page, not because someone promised more AI mentions.
Which schema types matter most for AI search?
There is no universal AI schema stack. Choose markup by page type and by the search features you are eligible for. Organization is useful for business details, Article for article metadata, and types such as Product, LocalBusiness, Event, VideoObject, or JobPosting for matching pages. Person can describe an author or profile, but it is not a documented E-E-A-T or AI citation shortcut.
Does FAQ schema still give you rich results in Google?
No Google FAQ rich-result benefit remains: the feature was retired on May 7, 2026. FAQPage remains a Schema.org type, but valid markup does not create a Google rich result or guarantee AI citations. Keep useful visible questions and answers for readers.
Can schema markup hurt my site?
Misleading or irrelevant markup can violate Google’s structured data guidelines and make a page ineligible for rich results. Mark up only content that is visible and true on the page, use the correct required properties, and keep dates, prices, ratings, and availability accurate. Valid syntax alone does not make a page eligible.
Should I use JSON-LD or microdata for schema?
Google supports JSON-LD, Microdata, and RDFa, and generally recommends JSON-LD because it is easier to add and maintain. Existing valid Microdata does not need to be replaced just for AI search. Pick the format your publishing system can keep accurate.
Do AI engines like ChatGPT actually read schema markup?
Major AI providers do not publish enough detail to claim that a specific schema type improves citation selection. Structured data may be present in retrieved HTML, but availability is not proof that a system uses it or gives it weight. Treat any citation effect as a hypothesis to test, not a documented benefit.
If you want a second set of eyes on your schema implementation, Book a free 30-minute call and bring the page you are reviewing.
