Guide Published 16 November 2026 11 min read

Schema Markup: Which Types Still Change Search Results

Google has cut more than a dozen rich result types since 2023. The structured data that still changes a listing, by type of site, where Merchant Center takes over, and how to tell whether the markup did anything.

Roman Makuev
Last updated

Schema markup is code, usually JSON-LD, that describes what a page contains in the vocabulary of schema.org: a product with a price, an event with a date, an organization with a logo. Search engines read it to understand the page, and Google uses some of it to show rich results, the stars, prices and thumbnails that make one result look different from the next. That last part keeps shrinking. Since 2023 Google has stopped showing about ten rich result types, FAQ among them, and the schema markup that still changes what a result looks like fits in one table.

Which rows of that table apply depends on what the site sells. A store, a local business and a publisher end up with three different short lists, and the long list that plugins add by default does little for any of them.

Schema markup, structured data, rich results: three different things

Schema.org is the vocabulary, with several hundred types, most of which no search engine displays. Structured data is that vocabulary applied to a page, in JSON-LD, Microdata or RDFa; Google recommends JSON-LD because it sits in one block and doesn’t have to be woven through the HTML. Schema markup is the everyday name for the same thing. A rich result is what Google decides to show for a page with structured data, and Google’s search gallery lists about two dozen features it supports. Valid markup makes a page eligible for one of them. Whether it appears is still Google’s call.

The ranking question has a plain answer. John Mueller, April 2025: “Structured data won’t make your site rank better.” It is used, he wrote, for displaying the search features in the gallery; markup outside those features is harmless and unlikely to change anything visible in Search. Google’s documentation adds that structured data helps it understand the content of a page. That leaves schema markup with two jobs worth paying for in Google: earning a rich result that changes how a listing looks, and feeding systems such as Merchant Center that read the same data.

What Google retired, 2023–2026

Most schema markup guides still list structured data types that no longer produce anything in Google. The removals came in waves, each announced as simplifying the results page.

FeatureWhat happenedWhen
FAQLimited to well-known government and health sitesAugust 2023
HowToRemoved, first from mobile results, then entirely2023
Sitelinks search boxRemoved from results; documentation removedNovember 2024
Course Info, Claim Review, Estimated Salary, Learning Video, Special Announcement, Vehicle ListingPhased out of results; removed from Search Console reports and the Rich Results Test in September 2025June 2025
Book ActionsAnnounced with the six above, later kept: Google removed the deprecation notice2025
Practice problemDeprecated; dropped from Search Console in January 2026November 2025
DatasetStill supported, but only by Dataset Search, not by Google Search resultsNovember 2025
FAQ, againNo longer shown for any site. Report and Rich Results Test support dropped in June 2026, Search Console API support in August 20267 May 2026

I read the list as a forecast. Most removals hit a feature that answered the query on the results page itself (a question, a salary, a course listing), and Google’s stated reason in 2025 was that the types weren’t commonly used. The features left standing mostly describe something a user is about to buy, book, watch or visit. That is no guarantee, and Google has never said it works to that rule, but it is where I would expect the next cuts to come from.

The types that still change search results

Sorted by the kind of site, the structured data that still earns a visible rich result is a short list. Each row has a condition, and the condition is usually where implementations fail.

SiteMarkupWhat it can changeCondition
EcommerceProduct (product snippet, merchant listing), ProductGroup for variants, return and shipping policyPrice, availability, rating, shipping and returns shown in the result; eligibility for shopping surfacesMarkup in the initial HTML, with values that match the page and the product feed
Local businessLocalBusiness, OrganizationBusiness details and logo that Google can use in its own panelsNo review stars for reviews of the business on its own site
Publisher or blogArticle, BreadcrumbList, VideoObjectTitle and image treatment in article features; the breadcrumb path; video thumbnails and key momentsBreadcrumbs show on desktop only since January 2025
Recipes, events, jobsRecipe, Event, JobPostingRecipe cards, event dates, the job search experienceEvery marked-up fact visible on the page
Reviews of third-party thingsReview, AggregateRating on a supported type (Product, Recipe, SoftwareApplication and others)Star ratings under the resultThe reviewed item isn’t the site’s own business
Forums, profiles, nicheDiscussionForumPosting, ProfilePage, Course list, Math solver, Education Q&A, Vacation rentalForum and profile treatments, carousels in narrow verticalsThe page is the thing the type describes

For a service business the honest schema markup list is short: Organization, LocalBusiness if there’s an address people visit, BreadcrumbList, Article on the blog, and VideoObject where there’s video. None of it produces stars. Stores have one choice to make inside Product. Google documents two variants: merchant listings, for pages where the customer can buy from you, and product snippets, for pages where they can’t, such as editorial reviews. A store’s product pages want the merchant listing properties, shipping and returns included, and those cover the product snippet as well. A plugin that adds FAQPage to every service page is now writing structured data that Google reads and does nothing with.

Check the results page before adding a type

The gallery says what Google can show. The results page for the query says whether it does. Before building Product or Recipe markup for a set of pages, search the queries those pages target and look at the listings already ranking. If competitors carry prices, stars or video thumbnails, schema markup can change how your listing compares. If nothing on the page carries the feature, the markup may be read and never displayed for that query, and the work is better spent elsewhere. It takes ten minutes and can save a full implementation for a feature the target queries never trigger.

Eligible isn’t shown: why markup stays invisible

“The Rich Results Test says it’s valid, so why are there no stars?” Validity is one of several conditions. Google’s guidelines add that the marked-up content has to be visible on the page, that the page has to meet its quality bar, and that some types carry their own restrictions. Markup that describes things a visitor can’t see on the page can also earn a structured data manual action. Google’s guidelines say what that costs: the page loses eligibility for rich results, and its ranking in web search isn’t affected. The Manual Actions report in Search Console is the first place to look when stars vanish from a whole site at once.

The restriction that catches most businesses dates from September 2019. Google stopped showing review stars for LocalBusiness and Organization when the reviews are of the business on the business’s own site, whether entered by hand or pulled in through a third-party widget. Google calls them self-serving. Google’s own FAQ on the change says the markup isn’t penalised and doesn’t need removing. It just never produces stars. Product reviews on a store are a different case: reviews of the products a site sells still qualify.

Review markup for LocalBusiness or Organization, placed on that business’s own site, is treated as self-serving: Google doesn’t show review stars for it (Google Search Central Blog, September 2019).

Breadcrumbs went the other way in January 2025, when Google stopped showing them in mobile results and started showing only the domain. BreadcrumbList structured data still shapes the desktop result and still describes the site’s structure. On a phone, the result looks the same with or without it.

Where schema markup stops deciding: Merchant Center

For stores, Product structured data has a second reader. Google’s shopping systems crawl product pages to check prices and availability, and they behave differently from Googlebot. In October 2024 Google added a warning to its documentation that Product markup generated with JavaScript can make Shopping crawls less frequent and less reliable, and recommended putting Product markup in the initial HTML. At Search Central Live in Zurich in December 2025 a Google engineer explained why: the shopping crawler visits often and fast, because it needs current prices and stock, and it doesn’t wait for scripts.

The feed is the other half. Merchant Center can take product data from a feed, from the schema markup on the page, or from both, and it compares them. With automatic item updates switched on, Merchant Center uses the structured data it finds on the site to correct prices and availability in Shopping listings, so the page markup feeds the listing directly. Checkout inside Google sits on top of that. Google announced its Universal Commerce Protocol in January 2026, and its Merchant Center help page for the native_commerce attribute says the attribute opts a product into the Buy flow in AI Mode and Gemini and decides whether the Buy button shows for that item. It lives in the product data source. Nothing in the page’s JSON-LD switches it on.

What ties it together is one source of truth. Price and availability appear in at least three places: the visible page, the structured data and the feed. When they disagree, Google has to pick one. Asked in July 2026 which source wins when stock status conflicts between a page and its feed, Mueller said there is no publicly defined order of precedence, and that conflicting metadata should be fixed. In a template, that means all three read from the same field in the same database.

Does schema markup help in AI search?

For Google’s own AI features, Google has answered in writing, twice. Its page on AI features and websites already said there was no special schema.org structured data to add, and its guide to generative AI features in Search, published in May 2026, says structured data isn’t required for generative AI search and that there is no special schema.org markup to add. It still recommends schema markup for rich results. Microsoft has said something different about its own systems: Fabrice Canel told SMX Munich in March 2025 that schema markup helps Microsoft’s LLMs understand content. I haven’t found documentation from OpenAI, Perplexity or Anthropic on whether their search features read structured data at all.

The only controlled test I know of is Ahrefs’, reported in May 2026: 1,885 pages that added JSON-LD, compared with pages that didn’t, over 30 days. AI Mode citations moved by +2.4%, ChatGPT by +2.2%, AI Overviews by −4.6%. The first two are within noise. The third was statistically significant, and Ahrefs said it couldn’t attribute it to the markup. Every page in the sample already had more than 100 AI Overview citations, so the study says nothing about pages that aren’t cited yet. That is the case people ask about, and I don’t have data on it. Nobody I’ve seen does. Until someone does, schema markup for AI visibility is a hypothesis, and it shouldn’t take budget away from the rich results it can be measured against.

What to do with markup Google no longer uses

Leave most of the schema markup in place. When Google cut FAQ and HowTo back in 2023, it wrote that structured data that’s not being used doesn’t cause problems for Search and has no visible effect, and it said the same when it removed the sitelinks search box. The June 2025 announcement noted that use of those types outside Google Search isn’t affected. Bing and other consumers can still read them.

Two things are worth doing anyway. Remove FAQPage where the FAQ itself was filler, because then the problem is the content. And check reporting: any dashboard that pulls FAQ rich result data from the Search Console API lost that field in August 2026 and may show empty rows or zeros, which a monthly report can easily read as a traffic drop. The FAQ filter disappeared from the Search appearance list in June 2026 as well, so year-over-year comparisons that included it now compare against nothing.

How to measure whether schema markup changed anything

Schema markup is one of the few SEO changes with a clean measurement path, because Google reports most rich results separately. In Search Console’s Performance report, the Search appearance filter splits clicks and impressions by feature: product snippets, merchant listings, review snippets, videos and others. Rolled out on a template, schema markup gives a natural test: add it to one group of pages, leave a comparable group without it, and compare the two over the same weeks.

QuestionWhere to lookWhen (our working windows)
Does Google read the markup?Search Console, the rich result report for the type: valid and invalid itemsDays after the pages are recrawled
Is the rich result shown?Performance, Search appearance, the feature’s filterTwo to four weeks
Did it change clicks?CTR for the test group against the control group, same queries, same periodFour to eight weeks
Did it change AI citations?No Google report; third-party citation trackers onlyUnknown; see the section above

The second row is where most expectations break. A report full of valid items with no impressions under the matching Search appearance filter means Google read the schema markup and chose not to show anything, and the reasons from the eligibility section above apply. The control group matters for the third row: rich results launch, change and disappear for whole verticals at once, and a before-and-after on a single page can’t separate the markup from the rest of the results page.

Building schema markup into the template

On most sites, schema markup belongs to the template, not to the page. A product template emits Product from the same fields that render the price and the stock badge. An article template emits Article with the real author and dates. One Organization block, with a stable @id, sits on every page and every other block refers to it. Built this way, the structured data can’t drift from the content, and drift is the usual reason a rich result that once showed stops showing.

CMS plugins are the other source of drift. Yoast, Rank Math and most themes emit their own schema markup, and a site that adds a second plugin or a hand-written block often ends up with two Organization nodes with different names or logos, or a FAQPage block on every page. Open the page source or the Rich Results Test on one page of each template and read what’s there before adding anything. Removing a duplicate is usually worth more than adding a type.

Two checks, because they answer different questions. Google’s Rich Results Test shows whether a page is eligible for the features Google supports, and ignores everything else. The Schema Markup Validator at validator.schema.org checks the vocabulary itself, including types Google doesn’t use. A page can pass the second and get nothing from the first.

Server-side JSON-LD is the safe default for schema markup, for the shopping crawler reason above and because markup that only exists after scripts run waits in Google’s rendering queue. Where a site renders on the client, the structured data is a separate line in the web development spec. Checking that the structured data still matches the page after each release belongs in the technical SEO audit, next to canonicals and status codes.