Schema Markup on 543 Top Homepages: Which Types Are Used

We fetched the homepages of the top 1,000 domains on the Tranco list on November 5, 2026 and read the HTML as a search crawler without JavaScript would see it. 543 of the 1,000 returned a readable HTML page. Among those 543, 285 homepages, or 52.5%, carry structured data in at least one of JSON-LD, Microdata or RDFa; 258, or 47.5%, use JSON-LD. The most common type is Organization, on 185 homepages, followed by WebSite on 171.

The number this page is careful never to print is "the top 1,000 sites". 412 domains returned an error, 41 disallowed crawling of their root in robots.txt, and 4 returned something other than HTML. Many of the top entries on a traffic-ranked list are content delivery networks, cloud endpoints, ad servers and trackers with no public homepage at all, and others block unfamiliar crawlers. So every percentage below has 543 under it, and the first section explains what the other 457 were.

This is a small measurement with a published method, a downloadable results file and a standard-library Python script, and it is kept deliberately separate from the much larger HTTP Archive Web Almanac figures it is compared against in section 06. For what each type is and when to use it, our schema types reference is the companion page.

Tranco ranks domains by combining popularity signals from several providers over a 30-day window. The list we used, JZPGY, was created on November 4, 2026 from data covering October 6 to November 4. Its top entries are not a list of websites people read; they are the domains that browsers and apps resolve most. Content delivery networks, cloud storage endpoints, analytics collectors and ad servers sit alongside the search engines and social platforms, and most of them have no homepage to crawl.

That shows in the outcomes. Of 1,000 attempts, 543 returned HTTP 2xx with an HTML content type after redirects. 412 returned an error: 278 of those were connection-level failures, recorded as URLError, which covers DNS resolution, TLS and refused connections; 75 were HTTP 403, 31 were 404, 7 timed out, 4 returned 429 and the remainder a scatter of other codes. 41 domains disallowed the root path for all user agents in robots.txt and were skipped without a fetch. 4 returned a non-HTML content type.

An example of why this matters for the headline: if the 412 error domains were counted as "no structured data", the any-schema share would read 29.8% and would be wrong, because most of those domains have no page on which schema could appear. The failure case we are avoiding is the one that makes a crawl of 1,000 names into a claim about 1,000 websites. We measured 543 homepages. The 403 and 429 responses are also a reminder that a crawler identifying itself honestly, as ours did, sees a different web from a browser.

Of the 543 reachable homepages, 285 carry at least one block of structured data in JSON-LD, Microdata or RDFa, which is 52.5%. The format split is lopsided: 258 homepages, 47.5%, have JSON-LD; 52, 9.6%, have Microdata; 6, 1.1%, have RDFa. 28 homepages carry both JSON-LD and Microdata, which is why the format shares add to more than the any-format share. 258 homepages, by coincidence the same number as the JSON-LD count, have none of the three.

Open Graph tags are counted separately because they are not schema.org structured data, though they are often lumped in. 316 homepages, 58.2%, carry an og:type meta tag, and 99 of those have Open Graph and nothing else. That is the clearest single pattern in the data: a meaningful group of large sites has done the work to control how a link unfurls in a social feed and has not described the site or organisation to search engines in any schema.org vocabulary.

Depth varies as much as presence. 141 homepages carry five or more distinct JSON-LD types, usually an Organization with nested ContactPoint and PostalAddress, a WebSite with a SearchAction, and a WebPage. Only one site produced a JSON-LD block our parser could not read. The example that illustrates the range: a homepage at rank 2 on the list carries Organization, SearchAction, WebPage and WebSite in one graph; a homepage at rank 9 carries nothing but Open Graph as served to us. The failure case is reading either as typical. The distribution is wide and the median site has either a small entity graph or nothing.

The table lists the 25 most common schema.org types across all three formats, counted once per homepage, with the JSON-LD-only count beside it. Organization leads on 185 homepages, 34.1% of the 543, and WebSite follows on 171, 31.5%. ImageObject at 135 and WebPage at 130 are mostly nested nodes inside those two. SearchAction at 110 and its companions EntryPoint and PropertyValueSpecification are the sitelinks search box pattern, still present on a fifth of top homepages although Google retired the visible search box in November 2024.

The two columns diverge in one place worth noticing. SiteNavigationElement appears on 18 homepages and on only 4 of them in JSON-LD; the other 14 carry it as Microdata, a holdover from the era when structured data was written into the HTML attributes of the navigation. WebPage has the next-largest gap between formats, 130 any-format against 117 JSON-LD, for the same reason: a handful of very large sites still mark their body element up with Microdata itemtype attributes rather than a JSON-LD block.

An example of how to read the counts: Question on 21 homepages, Answer on 20 and FAQPage on 20 are the same twenty-odd sites seen through three nodes of one pattern, not sixty sites. The failure case is summing the column. Types are counted per site, so each row is an independent share of 543 and the rows overlap heavily.

Splitting the 543 by Tranco rank produces the one result we did not expect. Among the 52 reachable homepages ranked 1 to 100, 20 carry structured data, 38.5%. Among the 75 ranked 101 to 250, 43 do, 57.3%. Ranks 251 to 500 give 59 of 126, 46.8%, and ranks 501 to 1,000 give 163 of 290, 56.2%. The band with the most traffic has the least markup.

The explanation is in who sits there. The top 100 is search engines, social platforms, operating-system vendors, app stores and the public faces of cloud providers. Several of the largest serve a homepage that is an application shell or a sign-in page rather than a marketing page, and several more serve a different page to an unfamiliar crawler than to a browser. In our crawl, homepages at ranks 2 and 12 carried JSON-LD entity graphs and rank 4 only an ItemList, while homepages at ranks 9, 10, 29, 34 and 38 carried none of the three formats in the HTML we received. Those are observations about what was served to one crawler on one day, not judgements about the sites.

An example of the right inference: in the 501 to 1,000 band, 135 of 290 reachable homepages, 46.6%, carry an Organization or WebSite node, so a site without either is in the slight majority of its tier, not an outlier, and adding one puts it with the better-marked half. The failure case is the opposite inference, that because two of the five reachable top-ten homepages serve no schema, schema does not matter for ranking. Those homepages are not competing for discovery in the way a brand site is. Reachable counts per band are small enough, 52 in the top band, that the percentages carry wide uncertainty, and we report them as what we observed rather than as estimates of a population.

Most discussion of schema markup is about the types that change how a result looks: product prices and ratings, FAQ dropdowns, article headlines and dates, local business hours, video thumbnails. On the 543 homepages we read, those types are uncommon. FAQPage is on 20, 3.7%. VideoObject is on 16, 2.9%. Article or NewsArticle is on 13, 2.4%. Product is on 10, 1.8%, with Offer on 23 because Offer also nests inside SoftwareApplication. LocalBusiness is on 4, 0.7%.

Two things explain the gap, and both are about what a homepage is. A homepage is rarely the page that should carry Product or Article markup; those belong on the product page and the article. Our crawl read one URL per domain and says nothing about inner pages, where the Web Almanac finds JSON-LD on 39% of desktop pages. The second is that the entity types, Organization and WebSite, are the ones Google's documentation has long associated with the homepage, and site owners have followed that guidance: entity markup is on a third of homepages, rich-result markup on a few percent.

An example of a correct use of these numbers: a publisher checking whether NewsMediaOrganization is unusual can see it on 22 of 543 top homepages, 4.1%, which for a publisher-specific type on a mixed list is a meaningful adoption. The failure case is treating FAQPage at 3.7% as evidence that FAQ markup is abandoned; a homepage sample cannot tell you what sites do on their help pages. Our rich results guide covers which types still earn a visible result and on which page they belong.

The nearest published reference is the HTTP Archive's Web Almanac, whose 2025 SEO chapter reports structured data across millions of home pages from a CrUX-based crawl with JavaScript rendered. Its 2025 figures for home pages are 50% with any structured data and 43% with JSON-LD, on both desktop and mobile, with Microdata on 17% of desktop and 16% of mobile home pages. It also reports that WebSite appears on 36% of mobile home pages, that WebSite and SearchAction are the two most common home-page types, and that only 2% of crawls had structured data added via JavaScript.

Our figures are 52.5% and 47.5% for any structured data and JSON-LD, and 9.6% for Microdata, on 543 unrendered homepages from a traffic-ranked top-1,000 list. The any-schema and JSON-LD shares land within a few points of the Almanac's; the Microdata share is about half, which is consistent with a top-1,000 list skewing to large, recently rebuilt sites that moved to JSON-LD. Our type ordering differs: Organization leads and WebSite follows, with SearchAction fifth, whereas the Almanac has WebSite and SearchAction first and second and Organization third. The Almanac's 2% figure for JavaScript-injected schema suggests our no-JavaScript parse undercounts by a small margin, though on a different population.

None of this is a validation of either dataset by the other. The Almanac measures millions of sites with a rendered crawl; we measured 543 with a raw fetch. The example of a fair sentence is "both find roughly half of home pages carrying structured data, with JSON-LD dominant". The failure case is averaging the two or describing our result as confirming the Almanac's.

Everything above can be re-derived from the results file and the script. The rows below state how the crawl was run and what it cannot tell you.

Download the 1,000-row results CSV (rank, domain, final URL, status, outcome, bytes, and the JSON-LD, Microdata, RDFa and Open Graph fields per domain) and the summary file with every count on this page. The crawler is a single standard-library Python script; the Tranco list is public under the id JZPGY.

Three things, and none of them is "add more schema". First, the common pattern among top homepages that carry markup is an Organization node with contact details, a WebSite node, and often a SearchAction, but only 42.9% of the 543 have an Organization or WebSite node at all. A site without them is in the majority; a site with them matches the better-marked group rather than leading it. Second, Open Graph without any schema.org markup is common enough, 99 of 543, that it is worth checking which of the two your own homepage actually has, because they are often conflated in audits.

Third, the rich-result types belong on the pages they describe. Product on 1.8% of homepages is not a sign that Product markup is rare; it is a sign that it lives on product pages, which this crawl did not visit. Our implementation guide covers placing each type on the right template. For teams asking whether any of this moves AI search citations, the controlled evidence is summarised in our AI Overview citation studies comparison, and the short answer is that the one test found no lift. The technical SEO work behind both pages is what our agentic SEO practice does for clients, on their own templates rather than a top-1,000 list.