Short answer
Schema markup labels the facts on a page in the schema.org vocabulary: who the business is, who wrote the page and what it sells. Google uses it to understand pages and to show rich results, and Microsoft says it helps Bing’s language models understand content. For AI Overviews and AI Mode, Google says no special markup is needed, and tests show chat assistants work from the visible text. Schema confirms what the page says; the content earns the citation.
Key takeaways
- Schema markup states a page’s facts in the schema.org vocabulary, used on more than 45 million web domains as of 2024, so machines read them without guessing.
- Google uses structured data to understand a page and the people and companies in it, and to make the page eligible for rich results, with no guarantee that either shows.
- For its AI features, Google says structured data is optional and there is no special schema.org markup to add; Microsoft says schema helps Bing’s language models understand content.
- In two small tests, chat assistants worked from the visible text: facts placed only in JSON-LD were missed, or read as plain page text.
- Use schema for identity and accuracy: Organization or Person with sameAs links, Article, Product and breadcrumbs, each matching what the page shows.
What schema markup is
Schema markup is code that labels the facts on a page with a shared vocabulary, so a machine reads them without guessing. The vocabulary is schema.org: as of 2024, more than 45 million web domains used it, with over 450 billion schema.org objects, and Google, Microsoft, Pinterest and Yandex are among the applications that read it [1].
It comes in three formats: JSON-LD, Microdata and RDFa. Google accepts all three and recommends JSON-LD, a block of code in a script tag that sits apart from the visible text and is the easiest to maintain at scale [2]. The types describe things: an Organization, a Person, an Article, a Product with its Offer, a BreadcrumbList for the path through a site.
What schema does for search engines
Google uses the structured data it finds to understand the content of a page and to gather information about the world, such as the people, books or companies in the markup [2]. Two uses matter most for AI search.
The first is identity. Organization markup on a home page helps Google understand a company’s administrative details and tell it apart from other organizations, and its sameAs property points to the brand’s profiles on other sites [4]. Article markup tells Google who wrote the page, and the author’s URL should lead to a page that identifies that person [5]. These are the facts an answer engine needs to resolve a brand or an author as one entity.
The second is eligibility. Structured data makes a page eligible for rich results, and Google states that correct markup enables a feature without guaranteeing that it appears [3].
Microsoft says more about its own systems. At the SMX Munich conference in March 2025, Bing’s Fabrice Canel confirmed that schema markup helps Microsoft’s large language models, the ones behind Copilot, understand content [6].
Where schema stops
Google is direct about its AI features. Its guide to generative AI search lists overfocusing on structured data among the common misconceptions: structured data is optional for generative AI search, and there is no special schema.org markup to add, though it stays useful for rich results [7]. Its page on AI features says the same for AI Overviews and AI Mode, and asks that structured data match the visible text on the page [8].
Chat assistants that read a page live show the same limit. In SearchVIU’s test of October 2025, ChatGPT, Claude, Gemini, Perplexity and Google’s AI Mode were asked for the prices on a test page; none of them reported a price that appeared only in JSON-LD, and ChatGPT found the three prices shown in the visible text [9]. In Mark Williams-Cook’s test of February 2026, ChatGPT and Perplexity repeated an address placed only inside invalid JSON-LD, which he read as the models taking in whatever the HTML contains [10]. The two tests differ in detail and agree on the point: neither showed an assistant treating schema as a special signal.
Rich results are shrinking too. Google has been phasing out lesser-used search features: breadcrumbs left mobile results in January 2025, and FAQ rich results stopped appearing in Google Search on 7 May 2026 [11]. And markup that describes content readers cannot see, or that misleads, breaks Google’s guidelines; a manual action for it costs the page its rich results [3].
Which schema types matter for AI search
The types worth the effort are the ones that state identity and the facts people ask assistants about.
| Type | What it states | Why it matters for AI search |
|---|---|---|
| Organization | The brand’s name, logo, address and sameAs links to its profiles | Ties the brand to the places that describe it, so it resolves as one entity [4] |
| Person | An author’s name, job title, employer and sameAs profiles | Connects each article to one identifiable expert [5] |
| Article or BlogPosting | The headline, the author, the publication date and the last update | States who wrote the page and how current it is [5] |
| Product and Offer | The product, its price and its availability | Keeps the price and stock in the markup the same as on the page [3] |
| BreadcrumbList | The page’s place in the site | Shows the structure; Google shows breadcrumbs on desktop only [11] |
| FAQPage | The questions and answers on the page | Describes the page for any system that reads schema.org; Google has shown no FAQ rich results since 7 May 2026 [11] |
How to add schema for AI search, step by step
- Start with identity: Organization markup on the home page, or Person markup for a personal brand, with sameAs links to every profile you control [4].
- Mark up every article with its author and dates, and point the author’s URL to a page about that person [5].
- Keep every important fact in the visible text as well: the price, the address, the author, the date. The visible text is what chat assistants read [9].
- Serve the JSON-LD in the HTML your server sends. Google can read markup that JavaScript adds later [2], while in SearchVIU’s test no assistant found JSON-LD added that way [9].
- Match the markup to the page, word for word for FAQs, and update both together [3].
- Validate with Google’s Rich Results Test and the Schema Markup Validator after every template change.
Steps 1 and 3 carry the most weight. Identity markup helps engines resolve who you are, and the visible text carries the facts they repeat.
What earns the citation
Schema is one part of entity coverage. The citation comes from the rest of it: pages that answer the questions behind a prompt, a brand described the same way across the web, and mentions on the sites assistants trust. I cover that work in entity SEO for AI search, and the wider discipline in what is answer engine optimization.
In my programs, schema sits inside that entity work. The tools I build cover entity and schema coverage and competitive analysis, so the gaps show against each competitor, and the fixes go into the site’s templates once, so every page carries them. That work was part of leading answer engine optimization (AEO) and generative engine optimization (GEO) for Calendly, one of the most recognizable SaaS platforms in the world. The answer engine optimization page shows how I run it for clients.
A worked example: an online language school
The school and its details here are illustrative. An online language school has Product markup for its three course plans with their prices, while its pricing page shows those prices only inside images. Asked what the courses cost, ChatGPT quotes a review site’s figures from the year before, two of them out of date. The school’s Organization markup has no sameAs links, and assistants mix it up with an older school of a similar name.
The fix takes a week. The prices go onto the pricing page as text, matching the Product markup. The Organization markup gains sameAs links to the school’s LinkedIn page, its business directory listings and its app store pages, and the About page states the same founding year and description. Within a few weeks, the answers the school tracks quote the prices from its own page, and the mix-up with the other school stops.
Frequently asked questions
Does schema markup help with AI search?
It helps search engines understand a page and the entities in it, and Microsoft says it helps Bing’s language models understand content. Google says its AI features need no special markup, and chat assistants read the visible text, so schema works alongside the content, and the content carries the facts.
Does ChatGPT read schema markup?
Tests differ in detail. In SearchVIU’s test of October 2025, ChatGPT missed every price that appeared only in JSON-LD; in a test of February 2026, it repeated an address found only in the markup, as if it were page text. Neither test shows ChatGPT treating schema as a special signal, so put every important fact in the visible text.
Do I need special schema for Google AI Overviews?
No. Google says there is no special schema.org structured data to add for AI Overviews or AI Mode. The structured data that supports rich results is worth keeping, as long as it matches the visible text.
Is FAQ schema still worth adding?
Google stopped showing FAQ rich results on 7 May 2026, so the markup no longer changes how a page looks in Google. It still describes the questions on the page for other systems that read schema.org, and it costs little where the questions and answers appear on the page word for word.
Which schema types matter most for AI search?
The identity types come first: Organization or Person with sameAs links to your profiles, and Article with its author and dates. Then Product and Offer for anything with a price, and BreadcrumbList for the site’s structure.
Sources
- Schema.org, home page, read 29 October 2026. schema.org
- Google Search Central, “Introduction to structured data markup in Google Search”, last updated 10 December 2025. developers.google.com
- Google Search Central, “General structured data guidelines”, last updated 10 July 2026. developers.google.com
- Google Search Central, “Organization (Organization) structured data”, last updated 8 September 2026. developers.google.com
- Google Search Central, “Article (Article, NewsArticle, BlogPosting) structured data”, last updated 8 September 2026. developers.google.com
- Search Engine Land, Barry Schwartz, “Microsoft Bing/Copilot use schema for its LLMs”, 20 March 2025. searchengineland.com
- Google Search Central, “Google’s guide to optimizing for generative AI features on Google Search”, last updated 10 July 2026. developers.google.com
- Google Search Central, “AI features and your website”, last updated 10 December 2025. developers.google.com
- SearchVIU, “Schema Markup and AI in 2025: What ChatGPT, Claude, Perplexity & Gemini Really See”, 2 December 2025. searchviu.com
- Search Engine Roundtable, Barry Schwartz, “Report: ChatGPT & Perplexity Treat Structured Data As Text On A Page”, 6 February 2026. seroundtable.com
- Google Search Central, “Latest documentation updates”, read 29 October 2026. developers.google.com