Does Schema Markup Help AI Citations?

By Mehmet, Top Seed AI · Published Oct 20, 2026

In short: No clearly documented test we found shows that adding schema markup makes AI assistants cite you more. Google says its AI features need no special markup, and the one such test with matched comparison pages, covering 1,885 pages, saw no clear citation gain. Schema still has documented uses, mainly Google rich-result eligibility and helping systems read your facts, so add it accurately, but not as a citation tactic.

Schema markup (structured data) is code on a page, usually JSON-LD, that labels facts such as a company name, an article's author or a product's price.

What schema markup is documented to do

Schema markup helps Google read a page and can make it eligible for enhanced Google results. It does not guarantee one.

Google's introduction says Google uses structured data to understand a page and to learn about things named in the markup, and that it can enable rich results. Its guidelines say correct markup does not guarantee a rich result. For homepage Organization markup, Google says it can help tell your organization apart from others in search results. Its gallery lists supported types such as Article, Organization and Product.

These are Google's statements about its own product, Google Search, not about ChatGPT, Perplexity or Claude.

What Google says about its AI features

Google says no special markup is needed to appear in AI Overviews or AI Mode.

Its guide to generative AI features (updated Jul 10, 2026) says structured data isn't required and there is "no special schema.org markup you need to add." It still suggests using markup as part of normal SEO, but ties the benefit to rich results. An older page (updated Dec 10, 2025) says the same and adds that markup should match the visible text.

This covers Google only, and not required does not mean no effect: neither page says whether markup influences which pages its AI features pick.

What Microsoft and the other assistants have said

A Microsoft blog says schema helps AI systems understand content, in general terms and without data. We found nothing on schema in the crawler pages of OpenAI, Perplexity or Anthropic.

A Microsoft Advertising blog post by a Bing product manager (Oct 8, 2025) says schema helps search engines and AI systems understand content. It does not say how Copilot uses it, gives no test data, and says nothing guarantees selection. Search Engine Land reported on Mar 20, 2025 that Microsoft's Fabrice Canel said at SMX Munich that schema helps Microsoft's language models understand content; that is secondhand, resting on a LinkedIn post Canel confirmed in the comments.

The pages in question are from OpenAI, Perplexity and Anthropic; they do not mention schema (checked Oct 20, 2026). That is silence, not evidence either way.

What tests have found

The only before-and-after test we found, with matched comparison pages, saw no meaningful citation gain from adding schema. Observational data show schema pages cited more, which does not prove cause.

Ahrefs, which sells SEO tools, first found that AI-cited pages were nearly three times as likely to carry JSON-LD across 6 million URLs. Because that gap could reflect site quality, it ran a second study (May 11, 2026; reported by Search Engine Journal). It followed 1,885 pages that added JSON-LD between August 2025 and March 2026 and compared them with matched pages that never did (Ahrefs states 4,000), counting citations for 30 days before and after with data from its own tools. Every page already had at least 100 Google AI Overview citations in February 2025, so the test says nothing about pages AI systems do not see yet. Against the comparison pages, citations changed by -4.6% in AI Overviews, +2.4% in AI Mode and +2.2% in ChatGPT. Ahrefs calls the AI Overviews decline statistically significant but small and unexplained, and the other two changes indistinguishable from zero. Ahrefs also notes that both groups were already declining before the change, and puts the AI Overviews gap at about 12 citations per page per day.

The limits matter. Owners added JSON-LD on their own, often changing other things too, so pages were not randomly assigned. All schema types were pooled, only JSON-LD present in the HTML was studied, and 30 days could miss slower effects.

The evidence is mixed, not empty. Schema pages are cited more in observational data, the before-and-after test found no gain, and two small tests suggest assistants do not use markup as structured data, though they disagree on whether its text is read.

As far as we found, no controlled test covers pages not yet cited, whether markup helps an assistant tell your company from a similarly named one, individual schema types or training data. A published response to the Ahrefs study raises the entity point; it is commentary, not data. We left out three other write-ups with mixed results (a conference slide deck and two company blog studies) that omitted key method details such as sample size, baseline or how schema was detected.

FAQ markup: what changed

Google no longer shows FAQ rich results in Search, so FAQ markup will not earn a Google result.

In August 2023 Google limited FAQ rich results to well-known, authoritative government and health sites. Its changelog says they stopped appearing from May 7, 2026 and that the FAQ documentation was removed on Jun 15, 2026; the old documentation address now opens the changelog. The 2023 post said there is no need to proactively remove FAQ markup: unused markup causes no problems for Search but has no visible effect there.

We found no test showing FAQ markup raises AI citations. AirOps's ChatGPT data lists FAQPage among higher-cited types (45.6%), the same observational association.

What is still worth doing

Add accurate markup for the basics as site hygiene, to help machines read your pages, not to change who AI assistants name.

A small, one-time effort is reasonable. Google ties markup to rich-result eligibility. Systems that read markup get a structured copy of key facts; a Microsoft blog says schema helps AI systems. Google says Organization markup helps tell organizations apart in Search; whether that carries over to AI assistants is unproven.

What to mark up:

• Organization on your homepage: name, logo, address, contact details.
• Article on posts and guides: headline, author, dates shown on the page.
• Product or Local business, only if it applies: price, availability or hours as shown.

Keep every marked-up fact in the visible text too. Google says markup should match it, and in the searchVIU test the facts that were found were visible.

How to check:

1. Open Google's Rich Results Test, enter a URL or paste code, and fix the errors it flags. Google describes it as showing which rich results a page's markup can generate; a pass does not guarantee one.
2. Open the Schema Markup Validator for a generic check of any schema.org markup, without Google-specific warnings. Google's testing page says the old Structured Data Testing Tool, minus its Google checks, now lives there.
3. Compare each marked-up fact with the visible page and fix anything outdated, missing or invented.
4. To test AI answers, ask a fixed set of buyer questions several times before and after, as in How to Check If ChatGPT Recommends Your Brand. Ahrefs suggests 5 to 10 pages with markup and 5 to 10 similar pages without, compared after 30 days; expect noise at that size. We sell measurement, so we benefit if you measure; these steps work by hand.

Neither tool says anything about AI assistants: one checks eligibility for Google's rich results, the other checks schema.org markup in general.

What not to do

Avoid markup that is untrue or invisible, and avoid expecting more than the evidence supports.

• Do not mark up content that is not on the page. Google says not to mark up content readers cannot see, or irrelevant or misleading content. A manual action removes rich-result eligibility but, Google says, does not change web ranking.
• Do not add fake or self-serving review markup. Google bars fake or undisclosed incentivized reviews, makes pages where a business controls its own reviews ineligible for star results, and says not to aggregate reviews from other sites (review rules).
• Do not add FAQ markup to win Google results.
• Do not treat schema as a citation lever, or pay for it as one. We found no evidence for that, and nobody can promise placement in an AI answer.

Questions people ask

Does schema markup help ChatGPT recommend my brand?

We found no evidence that it does. OpenAI's crawler page does not mention schema. In one small test ChatGPT missed a price that existed only in JSON-LD; in another it returned an address from made-up JSON-LD, apparently as plain text. AirOps found JSON-LD pages cited more in ChatGPT, an association, not proof of cause.

Is JSON-LD the right format for AI search?

Google recommends JSON-LD when your setup allows it and says Microdata and RDFa are equally fine if valid. Ahrefs tested only JSON-LD, and in the searchVIU test facts hidden in any of the three formats went unreported. No format has been shown to raise AI citations.

Do I need schema markup for AI Overviews?

Not according to Google. Its guide says structured data isn't required for generative AI search and no special markup is needed. Google still recommends markup as part of ordinary SEO because it helps with rich-result eligibility. Per its AI features page, a page does need to be indexed and eligible to appear with a snippet.

Should I remove FAQ schema now that Google dropped FAQ rich results?

Not necessarily. Google said in 2023 that unused structured data does not cause problems for Search, though it has no visible effect there, and we found no later advice to remove it. Keep it only if it matches visible questions and is easy to maintain.

Sources

• Google, Optimizing your website for generative AI features on Google Search, updated Jul 10, 2026: primary documentation.
• Google, AI Features and Your Website, updated Dec 10, 2025: primary documentation.
• Google, Intro to How Structured Data Markup Works, updated Dec 10, 2025: primary documentation.
• Google, General Structured Data Guidelines, updated Jul 10, 2026: primary documentation.
• Google, Structured Data Markup that Google Search Supports, updated Oct 16, 2026: primary documentation.
• Google, Organization Schema Markup, updated Sep 8, 2026: primary documentation.
• Google, Review Snippet structured data, updated Sep 8, 2026: primary documentation.
• Google Search Central Blog, Changes to HowTo and FAQ rich results, Aug 8, 2023 (with a Sep 14, 2023 update on How-to): primary announcement.
• Google, Latest Google Search documentation updates, entries dated May 8 and Jun 15, 2026: primary changelog.
• Google, Schema Markup Testing Tool (no date shown) and Rich Results Test (tool page, no date shown): primary tool descriptions. Schema Markup Validator: tool page, no date shown.
• Ahrefs, We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved., May 11, 2026 (modified Jun 9, 2026): vendor study, before-and-after with matched comparison pages, 1,885 pages, 30-day windows.
• Search Engine Journal, Schema Markup Didn't Move AI Citations In Ahrefs Test, May 11, 2026: press write-up of the Ahrefs study.
• AirOps, The Fan-Out Effect: What Happens Between a Query and a Citation, Apr 13, 2026: vendor study, observational, ChatGPT only, 353,799 pages.
• searchVIU, Schema Markup and AI in 2025: What ChatGPT, Claude, Perplexity & Gemini Really See, Dec 2, 2025 (test run Oct 30, 2025): vendor test, one made-up page, live fetching.
• Search Engine Roundtable, ChatGPT & Perplexity Treat Structured Data As Text On A Page, Feb 6, 2026: press write-up of one informal test.
• Gianluca Fiorelli, The Ahrefs Schema study is right. And it's testing the wrong thing, May 11, 2026: commentary, not a test.
• Microsoft Advertising, Optimizing Your Content for Inclusion in AI Search Answers, Oct 8, 2025: vendor guidance, no data.
• Search Engine Land, Microsoft Bing/Copilot use schema for its LLMs, Mar 20, 2025: press report, secondhand.
• OpenAI, Overview of OpenAI Crawlers; Perplexity, Perplexity Crawlers; Anthropic, Does Anthropic crawl data from the web, and how can site owners block the crawler? (page shows Apr 7, 2026): primary documentation, checked for any mention of schema.

External figures were checked against the linked pages on Oct 20, 2026. Where a study comes from a company that sells related tools, we say so. None of the outside studies measures your brand.

Related guides

• AI Visibility vs SEO: What's the Difference?
• How to Get Recommended by AI Assistants
• Why ChatGPT Doesn't Mention Your Brand

About this guide. Written by Mehmet, who runs Top Seed AI. We sell AI visibility measurement and improvement, so we have an interest in this topic; the sources above are there so you can check our reading. Published Oct 20, 2026.