The Number of Distinct Wikidata IDs Linked From Top-Site Homepage Markup Rose 284% Between 2023 and 2026

Summary

Picture a family-owned winery that gets a Wikidata entry after a few press mentions. The entry has its founding year, location and official website. The winery's own homepage, though, never says "that Wikidata item is us," so a machine has to guess whether the two refer to the same business.

Wikidata is the free, structured knowledge base that sits behind Wikipedia, and every entry has a stable ID such as Q42. Linking that ID from your Organization markup, through the sameAs property, is the most direct way to tie a website to a known entity.

Few sites do it, but the number is climbing fast. Across the top 1M sites, homepage markup pointed to 592 different Wikidata IDs in 2023 and 2,274 in 2026.

What we measured

We compared four October crawls from HTTP Archive (2023, 2024, 2025 and 2026), keeping the mobile homepages of sites ranked in the top 1M in each. For every homepage we parsed the rendered JSON-LD, flattened nested objects and @graph arrays, and pulled the sameAs URLs from Organization-type nodes.

A Wikidata link is any wikidata.org item URL with a Q number. Because one ID can be shared by many sites in the same network (a media group's regional sites, for example), we report both the number of homepages and the number of distinct IDs. The crawl's site list changes each year, so we compare shares as well as counts.

Distinct Wikidata IDs almost quadrupled in three years

The curve was flat for two years and then jumped.

Homepages linking Wikidata from any node, not only Organization nodes, rose from 1,781 to 3,377 over the same period (+89.6%). The homepage count grew more slowly than distinct IDs because in 2025 a single ID was shared by 468 homepages from one network; in 2026 the most shared ID appeared on 122.

The share grew at every rank level

Wikidata links are still uncommon, but more common than they were across the board.

Sites in the top 10,000 still lead by a wide margin, and the 100,001 to 1M tier rose from 0.6% to 1.0%.

Growth shows up across platforms

Counts here are homepages whose Organization markup links Wikidata, by technology detected on the page. A site can appear under more than one technology.

The rise is spread across custom JavaScript builds, open-source CMSs and hosted commerce. It coincides with growing interest in entity markup for AI search, though the data cannot show why individual sites added the links.

What this means

Search for your business on wikidata.org. If an item exists and it is accurate, add its URL (https://www.wikidata.org/wiki/Q followed by the number) to the sameAs list in your Organization markup, alongside your Wikipedia article if you have one. If the item has outdated details, fix them with sources. If no item exists, only create one if your business meets Wikidata's notability rules and you can cite independent sources. An AI search optimization agency should never create one for you without those sources.

Most of your competitors have not done this yet, which makes it a cheap way to stand out when AI engines decide which entity a website belongs to. To see how your brand is being identified right now, ask us for a free AI visibility audit.

Limits

The set of top-1M sites changes between crawls, and the 2023 crawl included more homepages (900,377) than 2026 (772,412). Shares are more comparable than raw counts.

Only homepages are measured.

A Wikidata link can point to an item about a parent company, a product or a place rather than the business itself; we did not check each item's subject.

Technology detection is HTTP Archive's and can miss or double-count stacks.

Frequently asked questions

Related studies

• Entity markup machines use to identify a brand97.8% of Brand Markup Skips Logo, Founder or Contact
• Entity markup machines use to identify a brandWikipedia Links in Brand Markup: Only 3.3% Have One
• Entity markup machines use to identify a brandProfilePage Markup Nearly Doubled Since Google Support
• Entity markup machines use to identify a brand1 in 10 Sites Mark Up a Brand Name Unlike Their Domain