Structured Data for AI Search: What Matters

Structured data helps describe the information on a page. It can identify the publisher, an article's author and dates, or the business behind a service. It does not guarantee a citation or a place in an AI answer.

Google says structured data is not required for its generative AI search features and there is no special schema.org markup to add. Its current AI optimization guide recommends keeping useful structured data as part of ordinary SEO, including eligibility for supported rich results.

For a business deciding where to spend its next hour, the practical question is whether a markup change will correct inaccurate information or support a relevant search feature. Adding more schema types simply to increase a score is a poor reason to make a change.

What should your structured data describe?

Choose markup that matches the page and facts that visitors can verify. These are useful places to start:

• Organization: the company's name, website, logo and relevant business details. Keep the company distinct from its founder. An Organization's sameAs should identify that organization; a founder's personal profile belongs with the Person. Google's Organization guidance explains those properties.
• Article or BlogPosting: the headline, actual author, original publication date and meaningful update date. These help Google understand article information. An author label alone does not establish expertise. Google's Article guidance covers recommended fields and author identification.
• Product: use it for an actual product and maintain accurate offers and availability where applicable. A consultancy service page should not be dressed up as a retail product just to chase a search feature.
• BreadcrumbList: describe the page's place in the site. Keep the visible navigation and destination URLs consistent.

Treat each field as a claim. If the page says a service is quoted after a consultation, do not invent a starting price in the markup. If a page identifies the company as its author, do not silently assign it to a named expert who did not write or review it.

Does FAQPage markup increase AI citations?

There is no sound basis in the platform documentation cited here for calling FAQPage the highest-impact schema type for ChatGPT, Claude, Perplexity or Google AI citations.

There is also a concrete change to Google's search features: Google stopped showing FAQ rich results on May 7, 2026, then removed the feature's documentation in June. The dates are recorded in Google's documentation updates.

Questions and answers can still help prospective customers. Explain what you deliver, what you need from the client and when your service is a suitable fit. Their value comes from answering real questions clearly. Do not promise extra Google search space or a citation advantage because you added FAQPage markup.

How are schema, robots.txt and llms.txt different?

These tools have separate purposes, and one cannot repair a problem in another.

Schema describes content. A JSON-LD block that identifies your organization will not make a blocked service page accessible.

Robots.txt expresses crawler preferences. OpenAI identifies OAI-SearchBot as its search crawler and GPTBot as its training crawler. Those controls are independent: a publisher can allow search while disallowing training. Hosting rules also matter, so check whether permitted crawlers can actually reach the page. OpenAI's crawler documentation lists the relevant user agents and IP ranges.

Llms.txt provides an overview and links for systems that use it. The llms.txt proposal describes a way to guide agents into a site's content. It is not a robots.txt replacement, an access-control mechanism or permission to disregard other restrictions.

For Google specifically, the AI optimization guide says Google Search ignores llms.txt and that maintaining it does not help or harm visibility or rankings. Maintain one if it serves a real use case, but do not let it distract from inaccessible pages or missing information.

An example from Rebel's own website

In a public audit of Rebel Online on September 7, 2026, our homepage's Organization markup put the founder's personal LinkedIn URL in the company's sameAs field. The same object already named Manuel Mendes as founder.

The correction is straightforward: place that personal profile with the founder and reserve the company's sameAs for verified company profiles. That makes the identity model more accurate. It does not prove that an AI service will recommend Rebel more often.

The same audit found that our blog's article cards were generated after a JavaScript request. The initial HTML contained no article links, while the sitemap listed the articles. Publishing the links in the initial HTML removes that dependency for readers or tools without rendering. Google can process JavaScript-generated anchors, so this finding alone does not prove a Google indexing failure. Google's link guidance explains the distinction.

These are different fixes: one improves the accuracy of an identity claim, and the other makes navigation available earlier. Neither needs a special AI schema.

How to audit a page without confusing readiness with results

1. Check access and eligibility. Inspect the HTTP response, robots rules, noindex and snippet controls, canonical URL and internal links. For Google AI features, review the site's inclusion setting in Search Console as well as indexing. Google's AI optimization guide describes the current requirements; eligibility never guarantees display.
2. Compare markup with the page. Confirm names, dates, URLs and claims against visible content and source records. Fix contradictions before adding optional fields.
3. Validate the relevant markup. Use the Schema.org validator for vocabulary and syntax checks, then Google's Rich Results Test for supported search features. A passing result cannot authenticate a testimonial or certify expertise.
4. Record observed visibility. Use a consistent set of buyer questions. Save the date, platform, exact prompt and cited URLs, and check whether the answer represents your service accurately. For a repeatable method, see our guide to testing AI visibility.
5. Track customer outcomes. Compare relevant search visits, audit completions and qualified enquiries over time. A citation is an observation; a new enquiry is a business outcome. A change in either does not, by itself, identify which website edit caused it.

Rebel's free AI Search Readiness Audit checks explicit technical signals on one public page. It does not establish whether that page is indexed or cited. For help prioritizing the findings alongside your content and commercial goals, see our AI Search Optimization service.