Does schema markup help AI search? What actually matters
Schema markup won't get you cited by ChatGPT or AI Overviews on its own, but a few types help engines understand who you are and what a page says. Here is what the evidence supports, the JSON-LD worth adding and how to check it is seen.
The seoscore team 9 min read
On this page
- What Google, Bing and AI assistants have said about structured data
- How answer engines use schema: entities, facts and trust
- Which schema types are worth adding for AEO
- FAQ schema after Google cut FAQ rich results: is it still worth it?
- JSON-LD examples you can copy
- Put schema in the HTML, not only in JavaScript
- How to validate your structured data
- What schema can't do
- Check your page with seoscore
- How to tell if adding schema changed anything
- Next steps
Schema markup helps AI search in a narrow way. It gives search engines a clear statement of who publishes a page, who wrote it, when it changed and what facts it contains, as long as it matches what the page visibly says. It does not earn citations on its own: Google says AI Overviews need no special markup, and independent tests suggest ChatGPT, Claude and Perplexity read pages as text rather than treating schema as structured data.
Schema is one signal in AEO (answer engine optimisation), the work of making pages that AI engines can read, trust and quote. Our guide to answer engine optimization covers the rest. This article sticks to markup: what the engines have said, which types are worth the effort and how to add them so they are actually seen.
What Google, Bing and AI assistants have said about structured data
Google's guide to optimizing your website for generative AI features lists "overfocusing on structured data" among the things you can ignore. It says structured data isn't required for generative AI search and there is no special schema.org markup you need to add. To appear in AI features, a page must be indexed and eligible to show in Google Search with a snippet.
That is not the same as "schema is useless". Google's page on AI features and your website still lists making sure your structured data matches the visible text on the page as a best practice, while saying you don't need new machine-readable files or markup to appear. Read together: normal, accurate structured data is fine and helps Google understand the page; there is no secret AI markup.
Microsoft Bing
Microsoft has been warmer. At SMX Munich in March 2025, Fabrice Canel of Bing said schema markup helps Microsoft's LLMs understand content, as reported by Search Engine Roundtable.
The current Bing Webmaster Guidelines, which now cover Copilot and grounding (the web sources an AI answer is based on and cites), are more careful. They say structured data may support clearer grounding but does not guarantee visibility or grounding traffic. They also say structured data must accurately represent visible content, and that misleading markup can be ignored or the site removed from the index.
ChatGPT, Claude and Perplexity
OpenAI, Anthropic and Perplexity don't document using on-page schema at all. The independent tests point one way:
- A searchVIU test placed eight prices for one fictional product in different places: visible text, JavaScript, JSON-LD only, hidden markup and more. No system reported the price that existed only in JSON-LD. ChatGPT named the visible prices and none from JavaScript or hidden markup.
- In a separate test reported by Search Engine Roundtable, Mark Williams-Cook put an address inside made-up JSON-LD, and ChatGPT and Perplexity found it. His reading was that the schema was being read as text, not used as structured data.
The tests disagree on whether ChatGPT picks up JSON-LD text, but agree on the important part: these assistants don't treat schema as a trusted data feed. If a fact matters, put it in the visible text.
How answer engines use schema: entities, facts and trust
An entity is a distinct thing a search engine can identify: your organisation, a person, a product, a place. Search engines build answers around entities, and schema is how a page says which ones it is about.
- Who you are. Organization markup with
sameAslinks to your official profiles (LinkedIn, Wikidata, Crunchbase, GitHub) ties your site to the same entity found elsewhere. That reduces the chance of being confused with a similarly named company. - Who wrote it. An
authorthat is a Person with a profile URL connects the content to a real, identifiable person. - When it changed.
datePublishedanddateModifiedstate freshness precisely instead of leaving the engine to guess. - Hard facts. Product, Offer and LocalBusiness pin down prices, availability, addresses and opening hours.
Consistency matters more than volume. If your name, logo, address and profiles match across your site, your markup and the rest of the web, engines can be confident they are all the same entity. That also supports GEO (generative engine optimisation), getting your brand named in AI answers, because assistants are more likely to name a brand they can identify clearly. Our GEO guide explains how that side works.
Which schema types are worth adding for AEO
Rich result status below follows Google's search gallery. Google retires features from time to time, so check it for the current list.
| Type | What it tells an answer engine | Google rich result | Priority |
|---|---|---|---|
| Organization | Your name, logo, official profiles via sameAs | Logo and organisation details | High |
| WebSite | Your site's name | Site name in results | Medium |
| Person (as author) | Who wrote the page and where to learn about them | None on its own | High for articles |
| Article | Headline, author, publish and update dates | Article features | High for articles |
| Product and Offer | Price, availability, ratings | Product snippets and merchant listings | High for shops |
| LocalBusiness | Address, phone, opening hours | Local business details | High for local firms |
| BreadcrumbList | Where the page sits in your site | Breadcrumb trail | Low to medium |
| FAQPage | Questions and answers on the page | Restricted since 2023 (see below) | Only where real Q&A exists |
| HowTo | Steps in a process | Removed in 2023 | Low |
Skip anything that describes content the page doesn't have. Markup for its own sake adds maintenance and risk, not visibility.
FAQ schema after Google cut FAQ rich results: is it still worth it?
In August 2023 Google limited FAQ rich results to well-known, authoritative government and health websites, so most sites no longer get the expandable questions in results. For everyone else, FAQPage markup is now only a description of the page.
It can still be worth adding, for a simple reason: question-and-answer content is exactly what answer engines quote. The markup itself is a minor part of that. What does the work is the visible block: a real question as a heading, followed by a direct answer in the first sentence or two. Mark that up with FAQPage and the markup and the content say the same thing.
It is clutter when the questions are invented to carry keywords, when the answers aren't visible on the page, or when the same FAQ block is pasted across dozens of pages. If nobody asks the question, leave it out.
JSON-LD examples you can copy
JSON-LD is the format Google recommends: a script block in the page that holds the markup separately from the visible HTML. Replace the example values with your own.
Organization with sameAs and logo
Put this on your home page, or site-wide.
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Organization",
"@id": "https://www.example.com/#organization",
"name": "Example Ltd",
"url": "https://www.example.com/",
"logo": "https://www.example.com/images/logo.png",
"description": "Example Ltd makes accounting software for UK sole traders.",
"sameAs": [
"https://www.linkedin.com/company/example",
"https://www.wikidata.org/wiki/Q000000",
"https://github.com/example"
]
}
</script>
@idgives the organisation a stable identifier that other blocks can point to, so you describe it once.sameAsshould list only profiles you control or that are definitely about you.- Google asks for a logo of at least 112 by 112 pixels that it can crawl.
Article with author and dates
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "How to file a self-assessment tax return",
"datePublished": "2026-03-02T09:00:00+00:00",
"dateModified": "2026-09-14T16:30:00+01:00",
"author": {
"@type": "Person",
"name": "Jane Smith",
"url": "https://www.example.com/team/jane-smith"
},
"publisher": { "@id": "https://www.example.com/#organization" },
"image": "https://www.example.com/images/self-assessment-16x9.jpg"
}
- The author URL should lead to a real profile page with that person's background.
- Change
dateModifiedonly when the content meaningfully changes, and show the same date on the page. - Google recommends images of at least 50,000 pixels (width times height).
FAQPage
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Do you deliver on Saturdays?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Yes. Orders placed by 2pm on Friday arrive on Saturday anywhere in mainland UK."
}
}
]
}
Each Question needs a name and an acceptedAnswer, and each Answer needs text. Both must match the visible question and answer word for word or close to it.
Put schema in the HTML, not only in JavaScript
Google can read JSON-LD that JavaScript adds to the page, because it renders pages. Most AI crawlers don't: they read the HTML your server sends and stop there. Markup added by a tag manager or a client-side script may never reach them. Our article on whether AI crawlers render JavaScript goes into the detail.
To check, look at the raw HTML rather than the browser's inspector, which shows the page after scripts run. Use view-source in your browser, or from a terminal:
curl -s https://www.example.com/your-page | grep -c "application/ld+json"
If that prints 0 but a browser inspector shows your JSON-LD, it is being injected by JavaScript. Move it into the server-rendered template or your CMS's SEO settings. While you are there, make sure the crawlers aren't blocked in the first place: see how to allow AI crawlers in robots.txt.
How to validate your structured data
- Run the URL through Google's Rich Results Test. It shows which rich result types Google detects and any errors or warnings in them.
- Run it through the Schema Markup Validator, which checks all schema.org types, including ones Google doesn't use for rich results.
- Compare what the tools found with the curl check above. A pass in the Rich Results Test doesn't prove a non-rendering crawler sees the markup.
- After publishing, watch the enhancement reports in Search Console for errors across the site.
The common problems:
- Markup that doesn't match the page. A price, rating or answer in the JSON-LD that isn't visible. Google and Bing both treat this as misleading.
- Missing properties. An Article with no author, or an FAQ answer with no text.
- Duplicate or conflicting Organization blocks. A theme, an SEO plugin and a tag manager each adding their own, with different names or logos. Keep one.
- Stale or fake dates. A
dateModifiedthat updates on every page load, or never updates at all.
What schema can't do
- It won't make thin content quotable. An engine quotes a clear, specific answer, not a well-labelled vague one.
- It won't override a robots.txt block or a
nosnippetrule. If the crawler can't read the page, the markup is irrelevant. - It won't replace the signals that do the heavy lifting: direct answers near the top of a section, cited sources, current content and a clear author.
- It won't create authority. Being cited and mentioned elsewhere is what makes an entity worth naming.
We have found no credible study showing schema causes more AI citations, and the published evidence comes from vendor tests rather than peer-reviewed research. Treat markup as a tidy, low-cost foundation, then spend most of your effort on the content. For the steps that matter most to Google's AI answers, see how to get cited in AI Overviews.
Check your page with seoscore
The free AEO and GEO checker runs our AEO checks on any URL. It tests entity markup and FAQ markup alongside the signals schema depends on: AI crawler access in robots.txt, whether the content is there without JavaScript, snippet rules, freshness, citations, direct answers and answer blocks. That tells you in one pass whether your markup exists and whether an AI crawler could reach it.
For the whole site, an SEO audit scores it out of 100, with AEO worth 15 of those points, and ranks every fix by the points it adds, so you can see whether markup is your biggest gap or a minor one. The SEO agent works through your audits each week and proposes fixes, including structured data, that change nothing until you approve them.
How to tell if adding schema changed anything
Record a baseline before you change anything, then compare after a few weeks. Engines need time to recrawl, and AI answers vary from one run to the next.
- Citations. AI mentions, on Starter and up with a free trial, reports where Google AI Overviews and ChatGPT cite your site, which pages, the questions asked and the other sites cited beside you. Run it before and after.
- Brand mentions. AI visibility asks ChatGPT, Claude, Gemini and Perplexity your customers' questions and scores how often, and how early, they name you.
- First-party reports. Google's AI optimisation guide points to generative AI performance data in Search Console, and Bing Webmaster Tools has an AI Performance report covering Copilot.
Don't read much into a single answer, and remember that if you changed content at the same time, you can't credit the markup alone.
Next steps
- Add one Organization block with your logo and
sameAsprofiles, and remove any duplicates. - Give every article a named author with a profile page, using Person markup.
- Add Article markup with honest
datePublishedanddateModifiedvalues that match the page. - Add FAQPage only where the page shows real questions with direct answers.
- Make sure all of it is in the server-sent HTML, then validate it.
- Measure citations and mentions before and after, and give it time.
Questions people ask
Is JSON-LD better than Microdata or RDFa for AI search?
Google supports all three but recommends JSON-LD because it sits in one block and is easier to maintain. There is no evidence that AI assistants prefer one format. What matters more is that the markup is in the HTML the server sends and matches the visible content.
Can schema markup get my site penalised?
Yes, if it is misleading. Google can apply a manual action for spammy structured data, and Bing's guidelines say misleading markup can be ignored or lead to removal from its index. Markup that accurately describes what is on the page carries no such risk.
Should I add llms.txt instead of schema markup?
They do different jobs and neither is required by Google for AI features. Schema describes entities and facts on each page, while llms.txt is a site-level file pointing AI tools to your key content. Our article on llms.txt explains what is known about whether assistants read it.
Does my WordPress SEO plugin already add schema?
Most popular SEO plugins add Organization or Person, WebSite, Article and BreadcrumbList markup automatically. Check the page source to see what is there, because a theme or second plugin can add a conflicting copy. Fill in the plugin's organisation, logo and social profile settings so the output is accurate.
Do I need a Wikidata entry for sameAs?
No. sameAs works with any authoritative profile that is clearly about you, such as LinkedIn, Crunchbase or an official social account. A Wikidata entry can help engines identify your entity, but only create one if your organisation meets Wikidata's notability rules.
Sources
- Google's Guide to Optimizing for Generative AI Features on Google Search developers.google.com
- AI Features and Your Website developers.google.com
- Schema Helps Microsoft's LLMs (Copilot) Understand Your Content www.seroundtable.com
- Webmaster Guidelines - Bing Webmaster Tools www.bing.com
- Schema markup and AI put to the test www.searchviu.com
- ChatGPT & Perplexity Treat Structured Data As Text On A Page www.seroundtable.com
- developers.google.com/search/docs/appearance/structured-data/search-gallery developers.google.com
- search.google.com/test/rich-results search.google.com
seoscore
Will AI answer engines quote you?
Run the free AEO checks on any page and see what to fix first.
Check my AEO