Does Schema Markup Help in AI Search? What the Research Shows

Schema markup is not the AI search shortcut many agencies are selling. Ahrefs tracked 1,885 pages that added schema and compared them with 4,000 similar pages that didn’t, and AI citations barely moved. Google now says plainly that no special schema is needed for its AI features. Schema still matters for rich results, for search engines that store it in their index, and for keeping your business facts consistent. It’s a supporting player, not the star.

I’ve built schema for ContentMender and every one of our clients, from Organization and LocalBusiness markup to Service and FAQ markup. I’ll keep doing it. But I’ve also watched agencies pitch “AI schema packages” as the fix for AI visibility, and the best research we have doesn’t support that pitch. Owners deserve a straight answer about where their money goes.

This guide covers what the controlled studies found, how AI tools actually read your pages, where schema still earns its keep, and how to set it up without falling for the hype.

Does Schema Markup Help You Get Cited in AI Search?

For pages that are already visible to AI tools, the best evidence says no meaningful lift. The most rigorous test so far found that adding schema didn’t increase AI citations on any platform it measured.

The first two rows explain why so many people get this wrong. Yes, AI-cited pages tend to have schema. But sites that bother with schema also tend to publish better content, earn more links, and fix their technical problems. When Ahrefs isolated the schema itself, using a method researchers call difference-in-differences, the effect disappeared.

Ahrefs summed it up well. If you’re already doing the rest of the SEO work, schema “isn’t going to be the unlock.” I agree, and I’d add one caveat that Ahrefs makes itself. Its study only looked at pages already cited heavily by AI. For pages AI tools have never seen, schema might still help them get understood in the first place. Nobody has proven that either way yet.

Can AI Tools Like ChatGPT Even Read Schema Markup?

When they fetch your page live, mostly not. Tests show AI chatbots read the visible text on your page and skip the hidden schema code.

The key is how each AI tool reaches your page. searchVIU describes four routes, which are training, indexing, search, and live fetching. When a chatbot fetches your page in real time, it reads what a visitor would see. When a search engine like Google or Bing crawls your page, it processes everything, schema included, and stores it. Google’s and Microsoft’s AI tools then pull from those indexes.

So the honest answer is that schema reaches AI through search engine indexes, not through the chatbot reading your code. That’s why the visible text on your page has to say the same things your schema says. If your business hours exist only in schema, a chatbot that fetches your page may never see them.

Where Does Schema Markup Still Earn Its Keep?

In traditional search, in the search engine indexes that feed AI tools, and in keeping your business facts consistent. Those benefits are real, and they’re why we still add schema to every client site.

The Otterly experiment is a useful example of that index benefit. After adding schema, its tracked pages saw AI Overview visibility rise sharply while ChatGPT visibility fell. That’s one SaaS website over three months, so I wouldn’t build a strategy on it. But it fits the pattern, since Google’s AI features draw on Google’s index and ChatGPT doesn’t.

For local businesses, the biggest structured data source isn’t on your website at all. BrightLocal’s study of 1.9 million local AI citations found Google Business Profile made up 28.5% of all citations across ChatGPT, AI Mode, and AI Overviews (BrightLocal, Sept 2026). Your profile is structured data that Google verifies, and it outweighs anything you add to your own code.

What Changed With FAQ Schema in 2026?

Google stopped showing FAQ rich results entirely. According to Google’s documentation changelog, the FAQ rich result feature stopped appearing in Google Search starting May 7, 2026, and Google removed its FAQ rich result documentation in June (Google Search Central updates).

That doesn’t mean FAQ sections are useless. It means the expandable FAQ dropdowns under search listings are gone. The FAQs themselves still do real work, for three reasons.

1. They answer real questions in plain text. AI tools pull passages that answer specific questions, and a clear question with a direct answer is easy to lift.
2. They can be cited from deep on a page. CXL’s analysis of 100 AI Overview citations found several citations from the bottom of pages came from FAQ blocks, since each question and answer stands on its own.
3. The markup still describes the page. FAQPage schema no longer earns a rich result, but it doesn’t hurt and it still labels the content clearly for search engines.

Our approach hasn’t changed much. We keep visible, click-to-open FAQ sections on our pages and our clients’ pages because they help readers. We keep the FAQPage markup because it matches what’s on the page. We just stopped promising anyone that FAQ schema will win them extra space in Google.

What Schema Mistakes Should You Avoid?

Most schema problems come from marking up things visitors can’t see, or from claims that don’t match the page.

• Hiding facts in schema only. Google says structured data should describe content users can see. AI tools that fetch your page live won’t see the hidden version anyway.
• Self-serving review stars. Google’s guidelines don’t allow review rich results for LocalBusiness or Organization markup when the business controls the reviews. Marking up your own testimonials won’t earn stars.
• Conflicting information. If your schema, website, and Google Business Profile list different phone numbers or hours, AI tools and customers both lose confidence.
• Buying an “AI schema package.” Google’s AI guide tells site owners to skip special AI markup, and the controlled data shows no citation lift for already-visible pages.
• Set-and-forget markup. Schema that still lists an old address or a closed location does more harm than good.
• Injecting schema with scripts only. Google can read JSON-LD added by JavaScript, but searchVIU’s test found most AI tools didn’t see it. Put core facts in the page HTML.

I’ve cleaned up all of these on client sites. The self-serving review markup is the most common one I find, usually added by a plugin years ago and never checked.

How Should You Use Schema Markup in 2026?

Treat schema as accurate labeling for what’s already on your pages, then spend most of your AI effort on content, reviews, and earned mentions.

Then check the work. Run your pages through Google’s Rich Results Test and the Schema Markup Validator, and make sure every fact in your markup appears on the page and matches your Google Business Profile.

If you want this handled, schema setup and cleanup are part of the technical work in our Hyperfocused Lead Generation and Hyperlocal Lead Generation programs. For where AI visibility really comes from, read why brand mentions matter more than backlinks in AI search and why updated pages get cited more in AI search. Or start with a free audit and we’ll show you what your schema says and whether it matches everything else.

FAQs