Schema markup in 2026: which structured data still matters for Google and AI search

In short: schema markup is structured data in the schema.org vocabulary, usually inserted as a JSON-LD block in the page code. It tells search engines exactly who the company is, where it is based and what a page offers. For Google it remains the basis for rich results such as prices, stars and breadcrumbs, even though Google has retired several types. For AI search it is not a lever: Google says no special markup is needed for AI Overviews and AI Mode, and a controlled study found no increase in citations. It still helps keep your company facts consistent.

We set up structured data on every website we build. This article shows which types are worth it for a business in Austria, what Google has dropped and how to check your markup in five minutes.

What is schema markup?

Schema.org is a shared vocabulary that Google, Microsoft and Yahoo launched in 2011; Yandex joined the same year. It defines types such as Organization, Product or Event, each with properties such as name, address or price. You add this information to the page code, usually as JSON-LD: a small script block that visitors don't see.

Google's guidelines are clear on one point: structured data must match the visible text. A price or rating that appears only in the code violates the guidelines and can lead to a manual action.

Which schema types does a business need?

For most service companies, Organization or LocalBusiness on the homepage, Service on service pages and BreadcrumbList everywhere is enough. Shops also need clean Product data, which is also the basis for listings in Google Shopping.

Which rich results does Google still show in 2026?

Google has been simplifying its results pages since 2025 and has retired many rich results. According to the Search Central changelog:

Still shown: product snippets with price and availability, merchant listings, review stars, events, recipes, videos, local business data, organisation logos and breadcrumbs on desktop. Google has also expanded some types, for example with shipping policies for merchants in November 2025.

The markup itself does not hurt if its rich result disappears. Google's notes on the FAQ removal say unused structured data causes no problems. You don't need to remove existing FAQPage markup.

Does schema markup help with ChatGPT and AI Overviews?

Probably less than many agencies claim. What we know as of October 2026:

• Google: In its guide to generative AI features (updated July 2026), Google writes that structured data isn't required for generative AI search and that there is no special schema.org markup to add.
• Ahrefs study (May 2026): 1,885 pages that added JSON-LD were compared with around 4,000 similar pages without schema. In ChatGPT and AI Mode, citations rose by about 2 %, which is not statistically significant. In AI Overviews they fell slightly. The pages studied were already well cited.
• Microsoft: Fabrice Canel from Bing said at SMX Munich in March 2025 that schema markup helps Microsoft's language models understand content. That applies to Bing and Copilot.

Our assessment: schema does not get you into AI answers on its own. What counts there are content that answers the question, consistent facts and mentions on other websites, as our guide Get cited by ChatGPT explains. Schema helps as a clean, machine-readable copy of your facts, and it costs little once it is set up.

How do you test your structured data?

1. Rich Results Test (search.google.com/test/rich-results): shows which rich results a page qualifies for and which errors Google sees.
2. Schema Markup Validator (validator.schema.org): checks all schema.org types, including those Google doesn't use for rich results.
3. Search Console: under “Enhancements” Google lists errors and warnings for every supported type across the whole site.
4. Compare with the page: does every price, address and rating in the markup also appear in the visible text?

Which mistakes do we see most often?

• Several plugins at once: theme, SEO plugin and shop each output their own Organization, with different names.
• Outdated data: an old address or old opening hours in the markup, while the website has been updated.
• Ratings without a source: five stars in the code with no visible reviews on the page.
• Everything as FAQ: every page gets FAQPage markup in the hope of extra space, which no longer exists.

Written byChristoph SpeiserProject lead & sales

Meet the team →