Does schema markup help AEO? What the evidence says

Schema markup has no demonstrated effect on AI citations. Google says structured data isn't required for generative AI search, and there's no special schema.org markup to add (opens in a new tab). When 1,885 pages that AI systems already cited added JSON-LD markup, their citations did not rise against matched pages that did not (opens in a new tab) (vendor study). Markup still helps pages become eligible for rich results in Google Search (opens in a new tab). Treat it as upkeep, not as a way to be cited.

Schema markup is one tactic within answer engine optimization (AEO), the work of making a company's pages and facts easy for AI answer engines to find, cite, and recommend. This guide separates three things: what platforms document, what studies measured, and what AEO HQ recommends. It supports the complete guide to answer engine optimization, and its sources were checked on September 27, 2026.

Scope and definitions

• Structured data, also called schema markup, is code in a page that labels the page's facts for software. Google describes it as "a standardized format for providing information about a page and classifying the page content" (opens in a new tab).
• Schema.org is the vocabulary most markup uses. It describes itself as "a collaborative, community activity" whose founding companies were Google, Microsoft, Yahoo, and Yandex (opens in a new tab).
• JSON-LD is a format that puts the markup in a script block, separate from the visible text. Google supports JSON-LD, Microdata, and RDFa, and recommends JSON-LD as the easiest to implement and maintain (opens in a new tab).
• A rich result is a search result with extra display elements, such as review stars or a price. Google says adding structured data "can enable" these results (opens in a new tab).
• A citation is a link to a source shown with an AI answer.
• Retrieval is the step in which an answer engine finds candidate pages. Grounding is basing the answer on the pages it retrieved.

This guide covers Google's AI Overviews and AI Mode, ChatGPT, Claude, Perplexity, and Microsoft Copilot. It does not cover product feeds, such as Google Merchant Center, which are a separate channel.

How schema markup works, and where it could matter

Markup could reach an AI answer by three routes. This framing is AEO HQ's, and the evidence differs for each:

1. At indexing. A search engine reads the markup when it processes the page, and an AI feature later draws on that index.
2. At a live fetch. An assistant opens the page while it answers and reads what the page contains.
3. In training. Markup is part of the text a model learns from. We found no study of this route.

What search engines say they do with markup

Google says it uses structured data "to understand the content of the page, as well as to gather information about the web and the world in general, such as information about the people, books, or companies that are included in the markup" (opens in a new tab). Its generative AI features retrieve pages from the Search index using its core ranking systems (opens in a new tab).

Google is equally direct that markup is not a requirement for those features. Its AI features page says no new markup is needed to appear in AI Overviews or AI Mode: "There's also no special schema.org structured data that you need to add" (opens in a new tab). Its optimization guide says structured data isn't required for generative AI search, but "it's a good idea to continue using it" because "it helps with being eligible for rich results on Google Search" (opens in a new tab).

Microsoft says more, with a limit attached:

• Bing's guidelines say structured data "may support clearer grounding but does not guarantee visibility or grounding traffic," and markup "must accurately reflect visible content" (opens in a new tab) (official documentation). This matters beyond Bing, because Microsoft Copilot sends the searches it generates to the Bing search service (opens in a new tab).
• A Microsoft Advertising blog post says schema "helps search engines and AI systems understand your content" (opens in a new tab) (platform guidance, no data).
• In March 2025, a trade publication reported that a Bing product manager said on stage that schema markup helps Microsoft's language models understand content (opens in a new tab) (secondhand report of a conference talk).

The crawler documentation from OpenAI (opens in a new tab), Anthropic (opens in a new tab), and Perplexity (opens in a new tab) does not mention structured data. None of the sources above reports a measured effect on citations.

What AI assistants read when they fetch a page

One practitioner test checked the live-fetch route. searchVIU built a page with eight prices, each exposed in a different way, and in October 2025 asked ChatGPT, Claude, Gemini, Perplexity, and Google AI Mode about it:

• No system found the price that existed only in JSON-LD, and hidden Microdata and RDFa were ignored as well (opens in a new tab).
• Only Gemini ran JavaScript during a live fetch, and AI Mode and Perplexity answered from their search indexes rather than fetching the page (opens in a new tab).

This was one page (practitioner test). It says nothing about the indexing route, and its authors write that markup "could very well be used" when search engines index pages (opens in a new tab). The practical rule holds either way: a fact that exists only in markup may never reach an assistant that reads the page.

Which rich results Google still shows

Rich results are the benefit Google documents most clearly, and the list has shrunk. Google's documentation changelog records:

• September 2023: HowTo rich results were removed, and FAQ rich results were limited to "well-known, authoritative government and health websites" (opens in a new tab).
• September 2025: course info, estimated salary, learning video, special announcement, and vehicle listing results were no longer shown (opens in a new tab).
• November 2025: practice problem and dataset markup were phased out of Google Search results (opens in a new tab).
• May 7, 2026: FAQ rich results stopped appearing in Google Search (opens in a new tab).

Google's list of supported features, last updated June 15, 2026, still includes Article, Breadcrumb, Organization, Product, Review snippet, Local business, Event, Job posting, Software app, and Video, among others (opens in a new tab). It has no entry for services. Schema.org defines a Service type as "A service provided by an organization" (opens in a new tab), but Google lists no search feature that uses it.

Two limits apply to every type. Google does not guarantee that structured data will show up in search results "even if your page is marked up correctly" (opens in a new tab). And a markup problem can lead to a manual action that removes a page's eligibility for rich results but "doesn't affect how the page ranks in Google web search" (opens in a new tab).

What studies of schema and AI citations found

The matched study

The strongest design so far comes from a vendor. Ahrefs tracked 1,885 pages that added JSON-LD between August 2025 and March 2026, compared them with about 4,000 matched control pages on other domains, and counted citations in the 30 days before and after (opens in a new tab). This is a difference-in-differences design: it compares the change in the pages that added markup with the change in similar pages that did not, which removes trends that affect every page.

The limits matter as much as the result:

• Every page already had 100 or more AI Overview citations in February 2025 (opens in a new tab), so the study says nothing about new sites or pages that are not yet cited.
• Schema types were pooled (Article, FAQ, Product, HowTo, Organization), the window was 30 days, and pages that add markup often change other things at the same time (opens in a new tab).
• The AI Overview decline was small in absolute terms: about 12 daily citations per page, in a sample where most pages got hundreds (opens in a new tab).
• The controls changed the answer. A simple before-and-after count showed AI Mode citations up 43%, but control pages gained almost as much (opens in a new tab).

Correlations

Cross-sectional data point the other way, and the sources themselves name the likely reason:

• In Ahrefs' first pass over 6 million URLs, pages cited by AI were almost three times as likely to have JSON-LD as pages that were not, and 53% of cited pages had schema (opens in a new tab). Ahrefs attributes the gap to sites that add structured data also investing in technical SEO, content, links, and upkeep (opens in a new tab).
• A 2025 preprint audited 1,100 URLs that Brave Summary, Google AI Overviews, and Perplexity cited for 70 product-intent prompts in B2B software (opens in a new tab) and reported a correlation of 0.63 between its structured data score and citation likelihood (opens in a new tab). Every audited page had been cited by at least one engine, so the comparison is among cited pages. The authors note that their "observational design may suffer from unobserved confounding" (opens in a new tab), and they are co-founders of Wrodium, a GEO company (opens in a new tab).

A correlation shows where markup is common. It does not show that adding markup causes citations.

A laboratory pipeline

SAGEO Arena, a peer-reviewed benchmark, built a search pipeline over about 170,000 web documents, with a keyword retriever that indexed titles, meta descriptions, headings, schema markup, and body text as separate fields (opens in a new tab). Rewriting those structural fields raised the retrieval hit rate by 22% (opens in a new tab). The same study found that the generator took most citations from body text, and structural fields were "cited less frequently despite its strong contribution to retrieval" (opens in a new tab).

Two cautions apply. The study did not test schema apart from titles and headings, and its authors say the pipeline does not fully reproduce commercial search engines (opens in a new tab).

What no study has measured

• Whether markup helps a new site, or a page that AI systems do not yet cite.
• Whether Organization markup or sameAs links change which companies assistants recommend. Google says Organization markup can help it "disambiguate your organization in search results" (opens in a new tab), but we found no study of an effect on AI answers.
• Whether one schema type helps more than another.

Steps

Steps 1 to 4 apply Google's and Bing's documentation. Steps 5 to 8 are AEO HQ's recommendations.

1. Put every fact in visible text first. Prices, scope, authors, and dates belong in the HTML that readers see. Google says not to add structured data about information that is not visible to the user, "even if the information is accurate" (opens in a new tab).
2. Mark up what the page is, using types Google documents. For most B2B sites that means Organization on the home page, Article with a named author on articles, and BreadcrumbList. For authors, Google asks for a URL that "uniquely identifies the author," such as a bio page, and accepts sameAs as an alternative (opens in a new tab). For the company, Google's sameAs property takes a URL such as "your organization's profile page on a social media or review site" (opens in a new tab). An example with placeholder values:

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Example Co",
  "url": "https://www.example.com",
  "logo": "https://www.example.com/logo.png",
  "sameAs": ["https://www.linkedin.com/company/example-co"]
}

3. Make the markup match the page. Google requires that structured data be "a true representation of the page content" (opens in a new tab) and lists making sure structured data matches the visible text among the SEO basics for its AI features (opens in a new tab). Bing says misleading markup "may be ignored and can affect trust and eligibility for enhanced features" (opens in a new tab).
4. Validate, then monitor. Google recommends the Rich Results Test during development and the rich result status reports after deployment (opens in a new tab).
5. Change the markup in the same edit as the page. When a price or a name changes, update both. Google's guidelines ask sites to provide up-to-date information (opens in a new tab).
6. Stop counting on FAQ markup. Keep visible FAQ sections where they answer real buyer questions. The markup no longer produces a rich result, as described above and in answer engine optimization antipatterns.
7. Test before you scale. To learn whether markup helps your own pages, add it to a few pages, leave similar pages unchanged, and compare both groups over the same period. Google suggests a before-and-after test on a few pages, recorded for a few months in the Performance report (opens in a new tab), and the matched study shows why a comparison group matters (opens in a new tab).
8. Spend the rest of the budget where evidence is stronger. That means specific facts on crawlable pages and coverage on other sites. See what replicates in AEO and GEO research and how to write content for answer engines.

The AEO checklist lists these checks with the rest of an audit. A related file, llms.txt, has a similar evidence record.

What the evidence shows and does not show

Antipatterns

An antipattern is a practice that looks helpful but fails or backfires. Longer entries are in answer engine optimization antipatterns.

1. Treating markup as the main citation lever. Citations did not rise after markup was added (opens in a new tab). Instead, keep markup as upkeep and spend the budget on facts and third-party coverage.
2. Putting facts only in JSON-LD. No assistant in one test read JSON-LD-only facts during a live fetch (opens in a new tab). Instead, state every fact in visible text.
3. Marking up what the page does not show. Google says not to mark up content that is not visible to readers of the page (opens in a new tab). Instead, mark up only what readers can see.
4. Adding FAQPage markup for a rich result. The FAQ rich result ended on May 7, 2026 (opens in a new tab). Instead, keep visible FAQs for readers.
5. Writing instructions to AI inside markup. Google defines spam to include attempts "to manipulate generative AI responses in Google Search" (opens in a new tab), and Bing says content designed to manipulate language models "may result in reduced visibility or removal from search experiences" (opens in a new tab). Instead, state facts plainly. The pattern is covered in generative engine optimization antipatterns.
6. Leaving old facts in the markup. An outdated price in markup contradicts the page. Instead, keep one fact sheet and update page and markup together. The planned entity and brand consistency checklist covers where facts drift.
7. Buying markup as a citation service. Google says third-party tools don't have access to its internal ranking data and "can't guarantee performance" (opens in a new tab), and it lists services "promising improvements for AI experiences and search formats (also known as 'AEO' or 'GEO' tools)" among those to evaluate critically (opens in a new tab). Instead, ask for the evidence behind any promised effect.

Checklist

FAQ

Does schema markup help with AI search?

No study has shown that adding it raises AI citations. In the best matched study, citations did not rise (opens in a new tab), and Google says there is no special schema.org markup to add for its AI features (opens in a new tab). Markup still supports rich results in Google (opens in a new tab), and Google says it helps it tell your organization apart from others (opens in a new tab) and understand who wrote an article (opens in a new tab).

Do ChatGPT, Claude, and Perplexity read schema markup?

Not when they fetch a page, in the one published test: no system read a price that existed only in JSON-LD (opens in a new tab). Their crawler documentation does not say whether their search indexes use markup. The search systems behind each assistant are described in how AI assistants find sources.

Should I still add FAQ schema?

Only if you want it for another reason. Google stopped showing FAQ rich results on May 7, 2026 (opens in a new tab). Keep visible FAQ sections that answer real questions, and mark up only questions and answers that readers can see. Markup cannot make a page a featured snippet either: asked how to mark a page as one, Google answers, "You can't" (opens in a new tab). See featured snippets and People Also Ask in the AI era.

Is there special schema for AI Overviews?

No. Google says "There's also no special schema.org structured data that you need to add" (opens in a new tab).

Which schema types should a B2B company use?

AEO HQ recommends Organization on the home page, Article with a Person author on articles, BreadcrumbList across the site, and Product with its offer on product pages. Google documents features for each of these (opens in a new tab). Schema.org has a Service type (opens in a new tab), but Google lists no feature for it, so add it only if it describes the page accurately.

Can schema markup hurt?

Wrong or misleading markup can lead to a manual action that removes rich result eligibility (opens in a new tab), and Bing may ignore it (opens in a new tab). In the matched study, AI Overview citations fell 4.6% against controls after markup was added, and Ahrefs could not attribute the decline to schema with confidence (opens in a new tab).

How do I test whether schema helps my site?

Add markup to a few pages, leave similar pages unchanged, and compare both groups over the same weeks. Count AI citations with a fixed set of prompts, run several times each, as described in our measurement design. A raw before-and-after count misleads: in the matched study, AI Mode citations rose 43% before controls were taken into account (opens in a new tab).

Change log

• September 28, 2026: First published.

Next steps

AEO HQ sells answer engine optimization services at fixed, published prices. The automated Instant AEO Audit ($499) records which structured data types your home page declares, along with checks of crawler access and llms.txt. Its limits are listed in our methodology.