Schema markup explained: what it does and doesn't do in 2026

Schema markup is code added to a web page — most often a single script tag written in JSON-LD — that labels the page's content using the shared Schema.org vocabulary. Instead of leaving a search engine to infer your business name, address, hours, and services from prose, the markup declares them as machine-readable facts: this string is our legal name, these numbers are our coordinates, this page describes a plumbing service in Salinas.

Three terms get used interchangeably, and the distinctions matter when you evaluate vendor claims. Structured data is the general concept — Google defines it as "a standardized format for providing information about a page and classifying the page content." Schema markup is the most common implementation of it, drawing on Schema.org's vocabulary of roughly 800 types. JSON-LD (JavaScript Object Notation for Linked Data) is the format that carries it: one self-contained block, separate from your visible HTML, that you can edit without touching the page layout. Google reads two older formats as well — Microdata and RDFa, which weave attributes into individual HTML tags — but its documentation recommends JSON-LD "if your site's setup allows it, as it's the easiest solution for website owners to implement and maintain at scale."

The vocabulary itself is neutral infrastructure. Google, Microsoft, Yahoo, and Yandex founded Schema.org in 2011 so the industry would share one way of describing things, and an open community process has maintained it since. Adoption is no longer a differentiator: Schema.org reports more than 45 million domains publishing over 450 billion Schema.org objects as of 2024, and Web Data Commons at the University of Mannheim found structured data on 52.1% of the 2.39 billion pages it parsed from the October 2024 Common Crawl. Adding markup will not make your site stand out. Going without it puts you in the minority.

No. Google has stated repeatedly that structured data is not a direct ranking factor: it changes how your listing appears in results, not where it ranks. Anyone selling schema as a rankings lever is selling against the vendor's own documentation.

The appearance change is the point. Rich results are enhanced listings — star ratings, breadcrumb trails, product prices, event dates — that Google may generate when a page's structured data qualifies it for a search appearance beyond the standard blue link. Eligibility is not a guarantee: Google decides per query whether to show one, and it can withdraw a rich-result type from the entire web, as the deprecation list below shows.

The measurable payoff shows up in click-through rate, and Google publishes the case studies itself. Rotten Tomatoes added structured data to 100,000 unique pages and measured 25% higher click-through on the enhanced pages. Nestlé measured an 82% higher click-through rate for pages that appear as rich results versus plain listings. Both numbers are self-reported results from motivated implementers, published by the vendor — but the direction matches what a richer listing at the same position should produce: more of the clicks that were already available.

The best evidence available in 2026 says no. Between August 2025 and March 2026, Ahrefs tracked 1,885 pages that added JSON-LD schema against 4,000 matched control pages that did not. Citations in Google AI Overviews fell 4.6% for the pages that added schema — a statistically significant drop — while the changes in Google AI Mode (+2.4%) and ChatGPT (+2.2%) were indistinguishable from zero.

The same study explains why so many tools claim otherwise. Across Ahrefs' six-million-URL dataset, AI-cited pages were almost three times more likely to carry JSON-LD than non-cited pages — a correlation the authors attribute to well-maintained sites, not to schema causing citations. Sites that add schema also tend to publish strong content, earn links, and load fast. When a vendor shows you a correlation chart as proof that markup wins AI citations, weigh it against this study, which controlled for exactly that.

Google says the same thing in plain language. Its guidance on AI features in Search states there is "no special schema.org structured data that you need to add" to appear in AI Overviews or AI Mode.

The work that does earn AI citations sits elsewhere: ranking in the classic index, answer-first structure, and content that arrives as server-rendered HTML. We cover those mechanics in our guide to Google AI Overviews.

Comprehension. A citation is a visible credit; comprehension is whether the machine has your facts right, and the two are measured on different scoreboards. In March 2025 at SMX Munich, Microsoft Bing principal product manager Fabrice Canel confirmed that schema markup helps Microsoft's LLMs — including Bing Copilot — understand web content. That statement is about parsing, not promotion, and it is the strongest direct endorsement any search engine has given schema in the AI era.

The clearest evidence that machine-readable structure changes machine accuracy comes from outside SEO entirely. A data.world benchmark (Sequeda, Allemang, and Jacob) had GPT-4 answer enterprise questions two ways: run against a raw SQL database, it scored 16.7% accuracy; grounded in a knowledge-graph representation of the same data, it scored 54.2% — three times the accuracy from structure alone. That is the honest version of the schema-and-AI story. Structure helps machines get facts right. No study yet shows it earning your pages more citations.

For a business, the practical target is Google's Knowledge Graph — its database of entities (people, organizations, places) and the relationships between them, the system behind knowledge panels. Google's documentation says Organization markup "can help Google better understand your organization's administrative details and disambiguate your organization in search results," and it influences which logo appears next to your name. Disambiguation is the working word: markup carrying your legal name, address, logo, and sameAs links (pointers to your verified profiles elsewhere on the web) tells Google exactly which real-world business you are, separating you from same-name businesses three states over.

The stakes rise with the sensitivity of the category. For a medical practice or a law firm, search engines apply higher trust thresholds, and an ambiguous entity record is a liability — reviews, credentials, and mentions scattered across the web never consolidate into one trusted profile. That is why entity work sits early in any serious healthcare marketing program, and why we treat schema as one input to AI visibility rather than the lever itself.

More than most owners realize, and the trend matters more than any single entry. Google has spent nearly three years retiring rich results it judged no longer worth showing. If a schema strategy was built around any of the following, the feature is gone no matter how clean the markup is:

• August 2023 — FAQ rich results restricted to "authoritative government and health websites." Most businesses lost them overnight, per the Google Search Central changelog.
• September 2023 — HowTo rich results deprecated for all sites.
• November 21, 2024 — the sitelinks search box removed globally after more than ten years, citing declining usage.
• June 12, 2025 — seven little-used types retired in one sweep, including Course Info, Estimated Salary, Learning Video, Special Announcement, and Vehicle Listing — types Google said "no longer added significant user value."
• May 7, 2026 — FAQ rich results stopped appearing in Search entirely, with the report and Rich Results Test support removed in June 2026.

Plenty still works in 2026: Product, Review and star ratings, Article, LocalBusiness, Organization, BreadcrumbList, Recipe, Video, and Event — the types most businesses need. The lesson from the removals is about posture. Keep the underlying content (FAQs remain useful to readers, and LLMs read them as plain text), implement the supported types, and never build a traffic projection on one rich-result feature. Google grants them per query and retires them by changelog.

Four earn their keep for most local and service businesses: Organization, LocalBusiness, BreadcrumbList, and Article. Two are situational, and two deserve a skip. Here is the per-type status as of July 2026, with rich-result standing drawn from Google's changelog and per-type documentation:

Two rows deserve a note. Service schema describes your offer, but no Service rich result exists — its value is entity clarity, which is why it sits at "optional" rather than "skip." And review stars are narrower than most owners assume: since 2019, Google has treated self-serving reviews — LocalBusiness or Organization markup carrying ratings of your own business on your own pages — as ineligible for stars. Ratings markup must describe another entity, so pasting your Google rating onto your homepage earns nothing.

Three routes, in rising order of control:

1. Your CMS or plugins. WordPress SEO plugins, Shopify themes, and most site builders already emit Organization, Article, and BreadcrumbList markup. Audit what exists before adding anything — duplicate or conflicting blocks are the most common mess we find in site audits.
2. Generate and paste. Schema generators build a JSON-LD block from a form; you paste it into the page head. Workable for a small site whose facts rarely change, fragile for anything that updates often.
3. Hand-written JSON-LD. For sites with a developer: one block per page, with the Organization declared once and referenced everywhere by its @id instead of redeclared. This is the approach Google's "maintain at scale" recommendation anticipates.

One rule keeps you safe. Markup must describe content visible on the page — marking up content users cannot see violates Google's guidelines and can draw a spammy-structured-markup manual action, which strips all your rich results at once. Honest mistakes carry no such penalty: unsupported types and deprecated markup left in place are simply ignored.

Test before and after publishing. Google's Rich Results Test shows which rich results a page qualifies for; the Schema.org Markup Validator catches syntax errors the Google tool ignores. After indexing, Google Search Console's enhancement reports track valid and broken items site-wide and report clicks on rich results. Expect a lag: changes surface days to weeks after Google recrawls the page.

Sequence it honestly. Schema is an afternoon of work that belongs near the start of a technical SEO pass, not a growth strategy on its own. Get the entity types right, add BreadcrumbList and Article, validate, then move to the work that moves rankings — content, links, and site quality. If you would rather hand off the whole pass, that is the shape of our SEO service: markup included, priced as the afternoon of work it is.