Schema.org for GEO: the markup you need for AI search

Schema.org is a vocabulary of structured data you use to tell search engines and AI systems exactly what’s on a page: where the company is, where the product is, who the author is, what the price is, and where the answer to a question lives. A person sees text; a machine sees only code, and markup translates one into the other. In 2026, the attitude toward it has shifted: Google removed a chunk of rich results, so people now add markup not for “stars in the results” but so search and ChatGPT and Perplexity understand the content correctly. Below we cover what still works, what changed, and how to implement markup without doing harm.

What Structured Data Is, in Plain Terms

Picture a page through a robot’s eyes: a wall of text with no hints about where the heading is, where the price is, or where the author’s name sits. Structured data adds those hints in machine language — “this is an organization,” “this is a product at this price,” “this is a question and answer.” The recommended format is JSON-LD: a small block of code in the page markup that’s invisible to the user but read cleanly by a machine. Markup doesn’t replace the content — it only explains what’s already on the page.

Why Markup in 2026 — and Why Not for Snippets

It’s worth being honest here, because there’s a lot of outdated advice around Schema. Google has steadily removed the snippets that used to be the main reason for markup: HowTo results disappeared in 2023, and FAQ snippets now show only for well-known government and health sites. In 2025, Google retired seven more structured-data types because they were barely used. So adding FAQ or HowTo markup “for the stars” no longer makes sense — those stars are gone. The real reason to add markup today is different: so a machine understands the page without guessing — both search and AI systems.

The paradox is that the value of markup hasn’t dropped over these years — it’s moved. It used to give a visible bonus in classic search; now it gives an invisible but no less important one: your page is understood more precisely and pulled into answers in places where the user never even sees ten blue links. That’s why we treat Schema not as a snippet trick but as basic hygiene for a modern site.

Which Markup Types Still Deliver

Some types still produce rich results in search, and some work “under the hood” — on understanding. Here’s what we rely on and who it’s for:

Schema and AI Search

This is where markup got a second life. AI systems synthesize an answer from sources they can easily understand, and clearly marked-up entities — organization, author, question-and-answer — help a model lift a ready fragment from your page and attribute it correctly. It’s not magic and not a guarantee of being cited, but it’s one of the cheapest ways to make content machine-readable. How this works together with the format of the text we covered in our piece on GEO and AI search: markup is the technical layer, the structure and authority of the text are the substance, and they only work together.

This is especially true for the entities AI likes to attribute: who the author is, which organization stands behind the material, which questions and answers are addressed. The more clearly that’s marked up, the less the model has to guess — and the lower the risk that you’re cited inaccurately or confused with someone else.

Where to Get Markup and Where to Put It

On most sites you don’t need to “write code” separately. SEO plugins (Rank Math or Yoast, for example) already generate basic Organization, Article, and BreadcrumbList markup automatically, and e-commerce platforms generate Product. The task is to avoid spawning duplicates: when the plugin, the theme, and hand-written code each add their own Organization block, the machine sees a conflict. So we keep a single source of truth for each type and write non-standard cases (special entities, complex products) by hand in JSON-LD. The markup block itself sits in the page code — the user doesn’t see it, but the crawler reads it on its visit.

Where to Start If You Have No Markup Yet

You don’t need to mark up everything at once. The order is simple: first, Organization or LocalBusiness across the whole site — the base that explains who you are. Then BreadcrumbList, because it almost always shows correctly in results and improves navigation for the crawler. Then the type that fits your business: Product for a store, Article for a blog. And only at the end, specialized types — if they genuinely exist on the pages. This order gives the most value for the least risk and doesn’t turn implementation into a month of work.

Common Markup Mistakes

On audits we regularly see the same misses with structured data:

1. Adding FAQ and HowTo markup “for the snippets” that Google no longer shows.
2. Marking up what isn’t on the page — a rating, a price, reviews — a direct guideline violation.
3. Using outdated microdata instead of the recommended JSON-LD.
4. Markup that contradicts the visible content: one price in the code, another on the page.
5. Duplicating or conflicting several Schema blocks on one page.
6. Not validating the markup and missing errors that cause it to be ignored.
7. Assuming markup alone will lift rankings without quality content underneath it.

How We Implement Markup

First, an audit: we look at what markup is already there, where it’s wrong or redundant, and which types are missing. Then we implement the needed types in JSON-LD: Organization for the whole site, Product for a store as part of e-commerce SEO, LocalBusiness for a local business (as in our piece on Google Business Profile), and Article for the blog. We handle the technical side carefully so the markup doesn’t conflict with the theme or cache — it’s part of the work on clean technical code. Finally, we validate everything and watch the enhancement reports in Search Console.

Markup isn’t a one-time job: the moment a platform, theme, or plugin updates, blocks can break or duplicate. So we periodically check the key templates (product, article, category) with a validator, so an error doesn’t sit unnoticed for months and quietly undermine the work.

The Takeaway

Markup isn’t about pretty snippets — it’s about a machine understanding your page without guessing.

Schema.org in 2026 is a tool for understanding, not cosmetics for the results page. Don’t chase snippets that are gone: add the types that actually work (Organization, Product, Article, BreadcrumbList), keep the markup matching the visible content, and use it as a technical bridge to AI search. If you’d like to check what’s going on with markup on your site — where it helps and where it adds risk — request an audit, and we’ll come back with a structured-data plan.

Mykhailo Umanenko

Project Manager · co-founder of heleum.studio

Co-founder and project manager at heleum.studio. 8+ years managing SEO and web projects: owns strategy, timelines, and results, and personally leads key accounts.