By Mehdi HassanDoes schema help AI citations?
No measurable effect has been shown. The clearest evidence is Ahrefs' controlled 2026 schema test, which found no meaningful lift in AI citations after adding structured data: AI Overview citations fell 4.6%, and citations in AI Mode and ChatGPT were flat. Google's own documentation says no special markup is needed for its AI features.
That runs against a lot of AEO advice, which still sells schema as a citation lever. The more useful framing is that schema is hygiene. It helps search engines understand what a page is and who published it, and some types still earn rich results. It doesn't turn a vague page into a quotable one. For the bigger picture of what does make content citable, start with our hub on how to structure content for AI search.
What did the Ahrefs schema test show?
Ahrefs ran a controlled 2026 test of whether adding schema markup changes how often pages are cited in AI answers. The result was no lift. AI Overview citations fell 4.6%, while AI Mode and ChatGPT showed no change either way.
- Change in AI Overview citations after adding schema (Ahrefs, 2026)
- -4.6%
- Change in AI Mode and ChatGPT citations in the same test
- Flat
A small drop like that shouldn't be read as "schema hurts." The sensible reading is that schema didn't help, and that whatever drives citations lies elsewhere. Like any single study, it has limits: one test, one period, one set of pages. But it lines up with what Google says publicly, and we haven't seen controlled evidence pointing the other way.
Why doesn't structured data drive AI citations?
Structured data doesn't drive AI citations because AI engines quote passages of text, and schema doesn't change the text. An AI answer is built by retrieving pages that seem relevant, then selecting sentences that answer the prompt. Markup can tell a crawler "this is an article by this author", but it can't make the answer on the page clearer, more specific or more trustworthy.
There's also a mismatch in what schema describes. Most types describe the page as an object: its type, author, date, breadcrumbs, product price. AI answers usually draw on the claims inside the page. The things that correlate with AI visibility sit outside markup altogether. In Ahrefs' study of 75,000 brands, web mentions correlated 0.664 with AI Overview visibility, against 0.218 for backlinks, and our guide to entity SEO for startups explains how to build those signals.
Which schema still earns rich results in 2026?
Several schema types still earn rich results or feed search features, and those are worth implementing for that reason alone. Google has narrowed the list over the last few years, so check its search gallery for the current set before building anything.
Schema type | What it does for a startup site | Worth adding? |
|---|---|---|
Organization | Clarifies your name, logo and official profiles (sameAs) | Yes, on the homepage |
Article (with author) | Clarifies headline, dates and named author | Yes, on every post |
BreadcrumbList | Shows the page's place in your site in results | Yes, site-wide |
Product and Offer | Price and availability in results for things you sell | Yes, if you sell products |
SoftwareApplication | App details and ratings for software products | Sometimes, if you meet the guidelines |
Review snippet | Star ratings for eligible items | Only with genuine, eligible reviews |
FAQPage | Rich results now limited to authoritative government and health sites | Optional hygiene; no rich result for most |
HowTo | Rich result removed by Google | No |
For marketing sites, the essentials are short: Organization, Article with a named author, and BreadcrumbList. Our guide to schema markup for websites walks through which types matter for a typical startup marketing site and how to implement them.
What changed with FAQ and HowTo rich results
In 2023 Google limited FAQ rich results to well-known, authoritative government and health websites, and stopped showing HowTo rich results. Many SEO plugins still add both by default, and older guides still recommend them as visibility wins.
FAQPage markup isn't harmful. It accurately describes a visible FAQ and may help parsers, so it's reasonable to keep. Just don't expect the expandable questions under your search result that sites used to get. The value of an FAQ now comes from the visible answers themselves, which is the subject of our guide to FAQ sections for AI search.
Google has continued to simplify search features since then, so treat any list of supported types, including ours, as something to check against the current documentation.
How should you implement schema on a startup site?
Implement schema once, in templates, generated from the same data that renders the page. Hand-pasted JSON-LD snippets drift out of date the moment someone edits the page, and inaccurate markup is worse than none.
Add Organization to the homepage
Name, URL, logo and sameAs links to your official profiles (LinkedIn, X, Crunchbase, GitHub). Keep the name identical everywhere.
Add Article to the post template
Headline, published and modified dates, and an author that links to a real author page.
Add BreadcrumbList site-wide
Generate it from your URL structure or CMS hierarchy so it never contradicts the visible breadcrumb.
Add type-specific markup only where eligible
Product, SoftwareApplication or reviews only when you meet Google's guidelines for each.
Validate and monitor
Run Google's Rich Results Test on each template and watch the enhancements reports in Search Console after launch.
In Next.js we render JSON-LD from the same CMS fields as the page, so dates, authors and titles can't disagree. Webflow and Framer support custom code in page settings, and Webflow can bind CMS fields into it. Either way, mark up only what's visible. A named author in markup should match a byline on the page, backed by an author page that builds trust.
What should you do instead to get cited by AI search?
To get cited, change the page, not the markup: put a direct answer at the top, make each section answer its own heading, use real tables for comparisons, show FAQ answers, and earn mentions on other sites. Those are the levers with evidence behind them or with Google's explicit endorsement.
- Answer first: reported analyses of ChatGPT citations say about 44% come from the first 30% of a page.
- Keep pages fresh: Ahrefs found AI-cited content is about 26% fresher than average.
- Earn mentions: they correlated with AI Overview visibility far more strongly than backlinks in Ahrefs' 75,000-brand study.
- Skip other shortcuts: llms.txt has no proven effect either; see whether your site needs llms.txt.
Then measure. Track AI referrals and run a regular prompt audit, as described in our guide to measuring AI search visibility, so you can see whether changes move anything. For Google's AI features specifically, our guide on how to get cited in AI Overviews collects the 2026 data.
