Adding schema markup has not been shown to get a page cited more often by AI answer engines. The best controlled study so far, Ahrefs’ May 2026 analysis of 1,885 pages that added JSON-LD, found no citation gain in ChatGPT or Google AI Mode and a small decline in AI Overviews. Google says structured data “isn’t required for generative AI search.” Microsoft says it “may support clearer grounding but does not guarantee visibility.” OpenAI, Perplexity and Anthropic don’t mention it in their documentation for site owners. Schema is still worth doing for four reasons that have nothing to do with chasing citations: the Google rich results that still exist, making clear which business or person a page is about, product and local business data, and honest markup that matches what the page says.
This guide sets out what each platform says, what the studies found, which Google rich results survived 2026, and the JSON-LD worth keeping. If you run a vacation rental site, copy-paste templates for your property pages are in Vacation Rental Schema Templates.
What Google, Microsoft, OpenAI, Perplexity and Anthropic say
| Platform | What its documentation says about structured data | Source |
|---|---|---|
| Google (AI Overviews, AI Mode) | Not required, no special markup; keep using it for rich results | AI optimization guide, updated July 10, 2026 |
| Microsoft (Bing, Copilot) | “May support clearer grounding” but guarantees nothing; must match visible content | Bing Webmaster Guidelines |
| OpenAI (ChatGPT) | Site-owner FAQ doesn’t mention it; shopping uses “structured metadata” from providers and product feeds | Shopping help |
| Perplexity | Crawler documentation doesn’t mention it | Perplexity crawlers |
| Anthropic (Claude) | Crawler documentation doesn’t mention it | Anthropic help center |
Google puts structured data in the “mythbusting” section of its guide to generative AI features, under “Overfocusing on structured data”: “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add. However, it’s a good idea to continue using it as part of your overall SEO strategy, as it helps with being eligible for rich results on Google Search.” Where the guide does point to structured inputs, it’s for commerce and local: AI responses “can include product listings, product information, and information about local businesses,” supplied through Merchant Center feeds and Google Business Profiles. The rest of Google’s guidance is in How to Optimize for Google AI Overviews and AI Mode.
Microsoft is the most positive of the five, within limits. Its Bing Webmaster Guidelines, which cover Bing search, Copilot and grounding API results, say: “Structured data may support clearer grounding but does not guarantee visibility or grounding traffic. Markup must accurately reflect visible content.” Bing’s markup help page says it uses annotations “to supplement and validate our data sources” and ignores markup that fails validation. After Microsoft’s Fabrice Canel spoke at SMX Munich in March 2025, attendee David Mihm wrote that Canel “confirms that schema markup helps Microsoft’s LLMs understand your content,” and Canel confirmed it in the comments (Search Engine Land, March 20, 2025). Mark Williams-Cook argues the quote in that report is about IndexNow and fresh content rather than schema, so treat it as a secondhand summary. A Microsoft Advertising post (Krishna Madhavan, October 8, 2025) says the same. Helping a system understand a page is not the same as getting it cited, and Microsoft doesn’t claim it is.
OpenAI’s publisher FAQ, as archived in August 2026, says “any public website can appear in ChatGPT search” and asks site owners not to block OAI-SearchBot; it doesn’t mention schema (archived copy). Shopping is different. OpenAI’s shopping help page says ChatGPT weighs “structured metadata from first-party and third-party providers (e.g., price, product description).” That data arrives through feeds and integrations: Shopify and Etsy catalogs are already integrated, and other merchants can apply for direct feeds, which OpenAI says “give you greater control over how your products appear.” More in How to Get Cited by ChatGPT.
Perplexity and Anthropic document their crawlers (PerplexityBot and Perplexity-User; ClaudeBot, Claude-User and Claude-SearchBot) and say nothing about structured data. What they ask of site owners is crawler access. See How to Rank in Perplexity AI and How to Get Cited by Claude.
The pattern: Google and Microsoft, which use structured data in their search indexes, describe it as an aid to understanding and rich results, never a requirement or a guarantee. The other three don’t mention on-page markup in their site-owner documentation at all; OpenAI’s structured product data comes from feeds.
What the studies show
Ahrefs’ controlled study (May 11, 2026). Ahrefs first analyzed 6 million URLs and found AI-cited pages were almost three times more likely to have JSON-LD than pages that weren’t cited; 53% of AI-cited pages had it. To test whether schema caused that, Louise Linehan and Xibeijia Guan identified 1,885 pages that added JSON-LD between August 2025 and March 2026, matched them with 4,000 control pages from other domains that had similar citation levels and never added it, and compared citations in the 30 days before and after. In their main difference-in-differences test:
| Surface | Change vs. matched controls | Ahrefs’ verdict |
|---|---|---|
| Google AI Overviews | −4.6% | Small but statistically significant decline |
| Google AI Mode | +2.4% | Statistically indistinguishable from zero |
| ChatGPT | +2.2% | Statistically indistinguishable from zero |
Three other tests pointed the same way. The caveats matter as much as the result: every page already had 100+ AI Overview citations, so the study can’t say whether schema helps pages that aren’t cited yet; all schema types were pooled; the window was 30 days; pages that add schema often change other things at the same time; and Ahrefs couldn’t explain the AI Overviews decline, so it doesn’t claim schema caused it. Its conclusion: “if the only reason you’re adding it is to get more AI citations on pages that are already visible, our data doesn’t support that bet.”
AccuraCast (September 17, 2025). AccuraCast ran over 2,000 prompts on each of ChatGPT, Google AI Overviews and Perplexity, collected about 9,000 cited sources, and found 81% of cited pages had schema markup. There was no comparison group of uncited pages, so this is a correlation. AccuraCast’s own conclusion was that schema “seems to matter to a very limited extent as a ranking, visibility, or citation factor.”
searchVIU’s fetch test (run October 30, 2025; published December 2, 2025). Michael Weber put one fictitious product on a test page in eight variants, each with its price in a different place: visible HTML, JavaScript, JSON-LD only, hidden Microdata, hidden RDFa and so on. He then asked ChatGPT, Claude, Gemini, Perplexity and Google AI Mode for the prices. “No system named the price that was only in the JSON-LD.” ChatGPT named the visible prices only; Gemini was the only one to pick up a JavaScript-inserted price on a direct fetch; Claude named no prices at all. searchVIU notes the test doesn’t show whether AI Overviews or Bing Copilot use schema stored in their indexes.
Mark Williams-Cook’s nonsense-schema test (reported February 2026). Williams-Cook put a fictional company’s address only inside made-up JSON-LD, with an invented @context, an invented type and invented properties, none of it valid schema.org. ChatGPT and Perplexity both returned the address. His reading: “as far as they were concerned, the JSON-LD was simply more text on the page.”
The two fetch tests disagree on whether text inside a JSON-LD block gets picked up. They agree on the point that matters: none of these assistants showed it was reading schema as schema, so a fact that exists only in markup may or may not reach an answer. Treat claims that pages with schema are some fixed percentage “more likely to be cited” with care: unless they come from a controlled comparison like Ahrefs’, they describe a correlation, like the 53% and 81% figures above. Our State of AI Citations 2026 benchmark reaches the same conclusion.
Google rich results in 2026: what’s left
Most of the “schema for AI” advice still in circulation was written for rich results that Google has since retired.
| Feature | Status | When |
|---|---|---|
| FAQ | Limited to “well-known, authoritative government and health websites,” then removed for everyone | August 2023; gone from Search May 7, 2026; documentation removed June 2026 |
| HowTo | Limited to desktop, then no longer shown | August 2023, then later in 2023 |
| Course info, estimated salary, learning video, special announcement, vehicle listing | Phased out | Announced June 12, 2025; documentation removed September 2025 |
| ClaimReview (fact check) | Phased out of Google Search results | Announced June 12, 2025 |
| Practice problems | Deprecated | November 2025 |
| Dataset | Being phased out of Google Search; still used by Dataset Search | November 2025 |
| Book actions | Kept, after an earlier deprecation notice was lifted | November 2025 |
Sources: Google’s documentation changelog, its August 2023 post and its June 2025 post, which says the change “won’t affect how pages are ranked.”
What remains: Google’s search gallery (updated October 16, 2026) still lists Article, Breadcrumb, Carousel, Course list, Dataset, Discussion forum, Education Q&A, Employer aggregate rating, Event, Image metadata, Job posting, Local business, Math solver, Movie, Organization, Product, Profile page, Q&A, Recipe, Review snippet, Software app, Speakable, Subscription and paywalled content, Vacation rental and Video. Google doesn’t guarantee any of them will show, even with valid markup.
Leftover markup isn’t a problem. When Google cut back FAQ results in 2023 it said “there’s no need to proactively remove it. Structured data that’s not being used does not cause problems for Search, but also has no visible effects in Google Search.”
What’s still worth doing, and why
1. Rich results that still exist. If your pages match a gallery feature (products, recipes, events, jobs, videos, reviews of other things), markup can make them eligible for a richer listing in classic results. Google’s structured data introduction cites case studies, including Rotten Tomatoes measuring 25% higher click-through on 100,000 pages with structured data and Nestlé measuring 82% higher click-through on pages that show as rich results. Those are Google’s own undated examples about classic search, not AI answers.
2. Making clear who you are. Google says Organization markup on your homepage “can help Google better understand your organization’s administrative details and disambiguate your organization in search results,” and can influence which logo appears in results and your knowledge panel (Organization documentation). Bing’s guidelines ask for “clear and consistent naming for people, organizations, products, and locations.” This is the most defensible link between schema and AI answers: Google says its AI features are “rooted in our core Search ranking and quality systems,” and Bing says Copilot relies on “the same core crawling, indexing, and ranking foundation” as its search, so what those engines understand about your organization carries into their AI features. Markup only helps if it agrees with your site and your profiles elsewhere. For the off-site side, see Entity SEO for AI Knowledge Graphs.
3. Product and local business data. Google says providing “both structured data on web pages and a Merchant Center feed maximizes your eligibility to experiences and helps Google correctly understand and verify your data,” and some experiences combine the two (product documentation). ChatGPT’s shopping results run on structured product metadata from feeds and providers. For a store, accurate Product and Offer markup, a Merchant Center feed and (where it applies) a ChatGPT merchant feed should all carry the same prices and stock. See AEO for Ecommerce. For a local business, LocalBusiness markup (name and address are required) sits alongside your Google Business Profile, which Google’s AI guide names directly; see Local Business AEO.
4. Honest markup that matches the visible text. Google’s general guidelines say not to “mark up content that is not visible to readers of the page” and warn that a structured data issue “can result in a manual action.” Bing lists misleading structured data among practices that can get markup ignored. Google also makes LocalBusiness and Organization pages ineligible for review stars when the business controls the reviews about itself (review snippet rules). Given the fetch tests, the practical rule is: every fact you want an AI answer to repeat (price, hours, address, author, date) goes in the visible HTML the server sends, and the markup mirrors it.
JSON-LD examples that validate
Google recommends JSON-LD as the easiest format to maintain but says all three formats (JSON-LD, Microdata and RDFa) are “equally fine” if valid. Bing has no preference. Put each block in a <script type="application/ld+json"> tag in the HTML your server sends. Google can read JSON-LD that JavaScript adds, but in searchVIU’s test only Gemini picked up a JavaScript-inserted price on a direct fetch, and no system read JSON-LD inserted by JavaScript. Each example below uses only schema.org types and properties, meets Google’s required properties for its feature, and uses example.com placeholders. Replace every value with what your page actually shows.
Organization, once, on the homepage or About page. Google lists no required properties.
{
"@context": "https://schema.org",
"@type": "Organization",
"@id": "https://www.example.com/#organization",
"name": "Example Co",
"legalName": "Example Co LLC",
"url": "https://www.example.com/",
"logo": "https://www.example.com/images/logo.png",
"description": "Example Co makes scheduling software for small veterinary clinics.",
"sameAs": [
"https://www.linkedin.com/company/example-co",
"https://www.crunchbase.com/organization/example-co"
],
"contactPoint": {
"@type": "ContactPoint",
"contactType": "customer support",
"email": "support@example.com",
"telephone": "+1-555-555-0100"
}
}
Only list sameAs profiles that exist, that you control and that describe the same organization.
BlogPosting (a subtype of Article) on each post. Google lists no required properties and recommends author, dates with a time zone, headline and image. Change dateModified only when the content really changes.
{
"@context": "https://schema.org",
"@type": "BlogPosting",
"headline": "How to Choose Scheduling Software for a Small Vet Clinic",
"description": "What to compare, what it costs and which features small clinics use.",
"image": ["https://www.example.com/images/vet-scheduling-16x9.jpg"],
"datePublished": "2026-03-02T08:00:00-05:00",
"dateModified": "2026-10-14T09:30:00-04:00",
"author": [
{
"@type": "Person",
"name": "Jane Smith",
"url": "https://www.example.com/team/jane-smith"
}
],
"publisher": {
"@type": "Organization",
"@id": "https://www.example.com/#organization",
"name": "Example Co"
},
"mainEntityOfPage": "https://www.example.com/blog/choose-vet-scheduling-software"
}
Product with an Offer on a page where people can buy. Google’s merchant listing feature requires name, image and offers, with a price and currency.
{
"@context": "https://schema.org",
"@type": "Product",
"name": "Trailhead Insulated Water Bottle, 32 oz",
"image": ["https://www.example.com/images/trailhead-32oz.jpg"],
"description": "Double-wall stainless steel water bottle with a leak-proof lid.",
"sku": "TH-32-BLK",
"brand": {
"@type": "Brand",
"name": "Example Outdoor"
},
"offers": {
"@type": "Offer",
"url": "https://www.example.com/products/trailhead-32oz",
"price": 34.00,
"priceCurrency": "USD",
"availability": "https://schema.org/InStock",
"itemCondition": "https://schema.org/NewCondition"
}
}
LocalBusiness, using the most specific subtype that fits (here Plumber). Google requires name and address.
{
"@context": "https://schema.org",
"@type": "Plumber",
"name": "Example Plumbing",
"url": "https://www.example.com/",
"image": "https://www.example.com/images/shop-front.jpg",
"telephone": "+1-555-555-0142",
"priceRange": "$$",
"address": {
"@type": "PostalAddress",
"streetAddress": "100 Main St",
"addressLocality": "Springfield",
"addressRegion": "IL",
"postalCode": "62701",
"addressCountry": "US"
},
"geo": {
"@type": "GeoCoordinates",
"latitude": 39.7990,
"longitude": -89.6440
},
"openingHoursSpecification": [
{
"@type": "OpeningHoursSpecification",
"dayOfWeek": ["Monday", "Tuesday", "Wednesday", "Thursday", "Friday"],
"opens": "08:00",
"closes": "17:00"
}
]
}
What to do with FAQ and HowTo markup now
Keep a visible question-and-answer section where readers have real follow-up questions; assistants that read the visible page can quote it. The FAQPage markup on top is optional: it no longer earns anything in Google, no AI platform documents using it, and Ahrefs’ study (which pooled FAQ markup with other types) found no lift. Existing markup can stay, since Google says unused structured data does no harm, but don’t add it expecting a result, and never mark up questions that aren’t on the page. The same goes for HowTo: write the steps out clearly and skip the markup.
How to audit your structured data
- Run your key templates (homepage, article, product, location) through Google’s Rich Results Test for Google features and the Schema Markup Validator for general schema.org validity.
- Check Search Console’s rich result reports for errors across the site.
- View the page source, not the rendered page, and confirm the JSON-LD and the facts it describes are in the HTML the server sends.
- Compare every marked-up value with the visible page: prices, stock, hours, addresses, authors, dates and ratings.
- Confirm
dateModifiedchanges only with real edits, not on every deploy. - Open each
sameAsURL and make sure it’s live and describes you. - Look for duplicates: a theme or plugin and a hand-added block both describing the same page in different ways.
How to test whether schema changes anything for you
If you want your own answer, copy Ahrefs’ design at small scale: pick 5 to 10 pages to get new or fixed markup and 5 to 10 similar pages that won’t, change nothing else on either group, and compare them after 30 days or longer. Track appearance rate (the share of answers that name you) and citation rate (the share that link to your site) for a fixed set of prompts on each engine, run repeatedly, because AI answers vary from run to run. If both groups move together, it was the platform, not the markup. Metric definitions and sampling are covered in How to Measure AI Citations, and the wider list of what matters more than markup is in The AI Citation Readiness Checklist.
Prefer to self-assess first? Download the 25-item AI Citation Readiness Scorecard (PDF), a printable self-assessment across schema, content architecture, authority, measurement and conversion. Free, instant download, no email required.
Want it done for you? The free AI Visibility Snapshot tests 10 prompts on ChatGPT, Google AI Overviews and Perplexity, with a short note on where you’re cited and where you’re not. The Citation Audit tests 30–50 locked prompts across six surfaces (ChatGPT, Google AI Overviews, Perplexity, Gemini, Claude and the OpenAI Responses API) and delivers competitor gaps, schema and content fixes and a prioritized 60–90 day roadmap in 5 business days. It’s $1,500, fully credited toward an Implementation Sprint or retainer signed within 30 days. We don’t guarantee citations or rankings. See pricing.
Request a free Snapshot or a Citation Audit →
FAQ
Does schema markup help a page get cited by AI?
Not on current evidence. In Ahrefs’ May 2026 study of 1,885 already-cited pages that added JSON-LD, ChatGPT and AI Mode citations didn’t change measurably compared with matched controls, and AI Overview citations fell 4.6%. Google says structured data isn’t required for its AI features. Keep schema for rich results and entity clarity, and put the facts you want cited in the visible text.
Do ChatGPT, Perplexity and Claude read JSON-LD?
Not as structured data, as far as published tests show. In searchVIU’s October 2025 test, none of five AI systems named a price that existed only in JSON-LD. In Mark Williams-Cook’s 2026 test, ChatGPT and Perplexity returned an address from deliberately invalid JSON-LD, which suggests they read the block as plain text. OpenAI, Perplexity and Anthropic don’t mention schema in their site-owner documentation.
Does Bing or Copilot use structured data?
Bing does, as an input to its index. Its Webmaster Guidelines say structured data “may support clearer grounding but does not guarantee visibility or grounding traffic,” and that markup must match visible content. An attendee’s report from SMX Munich in March 2025, which Microsoft’s Fabrice Canel confirmed in a LinkedIn comment, says schema helps Microsoft’s LLMs understand content. That’s an aid to understanding, not a promise of citations.
Is FAQ schema still worth adding?
Not for Google. FAQ rich results stopped appearing on May 7, 2026, after being limited to government and health sites in 2023. A visible FAQ can still help readers and be quoted by assistants that read the page. The markup is optional, and existing FAQPage markup doesn’t need to be removed.
Which schema types are still worth implementing?
Organization on your homepage, Article or BlogPosting on editorial pages, BreadcrumbList for site structure, Product and Offer on pages that sell something, and LocalBusiness for a physical business. Add other types only when your page matches a feature in Google’s search gallery, such as Event, Recipe, Video or Job posting.
JSON-LD, Microdata or RDFa?
JSON-LD, because it’s the easiest to maintain. Google recommends it but says all three formats are equally fine if valid, and Bing says it doesn’t prefer one. Whatever the format, serve it in the page’s HTML rather than relying on JavaScript, and keep it consistent with the visible text.
Sources:
- Google Search Central. Optimizing your website for generative AI features on Google Search. Last updated July 10, 2026. developers.google.com.
- Google Search Central. Latest Google Search documentation updates (FAQ deprecation May 2026 and removal June 2026; 2025 structured data retirements). developers.google.com.
- Google Search Central Blog (August 8, 2023). Changes to HowTo and FAQ rich results. developers.google.com.
- Google Search Central Blog (June 12, 2025). Simplifying the search results page. developers.google.com.
- Google Search Central. Structured data markup that Google Search supports (search gallery). Last updated October 16, 2026. developers.google.com.
- Google Search Central. Introduction to structured data markup in Google Search. developers.google.com.
- Google Search Central. General structured data guidelines. developers.google.com.
- Google Search Central. Feature documentation for Organization, Article, Product, merchant listings, local business, review snippets and structured data generated with JavaScript.
- Microsoft Bing. Bing Webmaster Guidelines. bing.com.
- Microsoft Bing. Marking up your site with structured data. bing.com.
- Madhavan, K. (October 8, 2025). Optimizing your content for inclusion in AI search answers. Microsoft Advertising. about.ads.microsoft.com.
- Schwartz, B. (March 20, 2025). Microsoft Bing/Copilot use schema for its LLMs. Search Engine Land. searchengineland.com.
- OpenAI Help Center. Publishers and developers FAQ. Archived August 27, 2026. web.archive.org.
- OpenAI Help Center. Shopping with ChatGPT search. help.openai.com.
- OpenAI. Power product discovery in ChatGPT (merchants). chatgpt.com.
- Perplexity. Perplexity crawlers. docs.perplexity.ai.
- Anthropic. Does Anthropic crawl data from the web, and how can site owners block the crawler? support.claude.com.
- Linehan, L. & Guan, X. (May 11, 2026). We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved. Ahrefs. ahrefs.com.
- Divecha, F. & Bergerou, J. (September 17, 2025). Does Schema Markup Increase Generative Search Visibility? AccuraCast. accuracast.com.
- Weber, M. (December 2, 2025). Schema markup and AI put to the test. searchVIU. searchviu.com.
- Williams-Cook, M. (May 28, 2026). Schema, LLMs and the Low Bar for “Evidence” in GEO. markwilliamscook.substack.com. First reported by Search Engine Roundtable, February 6, 2026.
All sources checked November 10, 2026.
Related reading: Vacation Rental Schema Templates, Entity SEO: How to Get Your Brand into Knowledge Graphs and AI Answers, How to Optimize for Google AI Overviews and AI Mode, The AI Citation Readiness Checklist, How to Measure AI Citations.