Download websites that use Open Graph

About Open Graph

Open Graph markup is the part of a page written for a reader who never visits it. The tags tell a social network or a messaging app which title, image and description to show when a link is pasted somewhere else, and a site that fills those fields has thought about what happens to its pages away from its own domain. The thinking is the signal. The tags themselves are trivial to add.

Meta introduced the protocol in 2010 to make pages behave like objects in a social graph, published it on Open Source terms, and then watched every other platform, competitors included, build its link previews on the same fields. As of September 2026 we find the markup on 49,950,141 sites. The Miscellaneous label we file it under is slightly misleading, since search ranking is not what the tags do, but it reflects where the work usually sits inside an organisation.

The commercial reading has changed as adoption spread. In the early years, presence separated organisations with a marketing function from organisations without one. Today a content management system, a store platform or a search plugin emits the tags with no human involvement, so presence has become a weak positive while absence has become a strong negative. A site with no preview metadata at all is either very old, hand built by somebody uninterested in distribution, or not really a publishing surface in the first place.

Used carefully the column still does real work, because the useful comparisons are relative rather than absolute. Within a market segment, the sites missing the markup are the ones whose links look broken when shared, which is a concrete opening for anyone selling web work. Within a portfolio, missing tags flag the pages that leak attention every time somebody posts them. Across an industry, the presence rate is a rough index of how much of that industry treats its website as a distribution channel rather than a brochure that happens to be online.

United States accounts for 26,530,914 of the sites carrying the markup, with the suffix weight on .com. The distribution mostly repeats the distribution of hosting, and our country attribution follows serving infrastructure rather than the owner's address, so the column is a starting point for segmentation and not a conclusion about which markets care more about being shared. Suffix, page language and the contact details on the site itself are the corrections to apply before any national claim is made.

The limits are simple to state. We record that the tags exist on the page we fetched, not that they are correct, not that the image they point at loads, and not that their values match what a visitor actually sees. A site can carry the full set and still produce an empty preview, and we cannot tell you that from the file. Checking one domain for the tags takes a few seconds in any browser, which is precisely why the per domain question is not the one worth paying for. The download answers the population version instead. Which domains in a country, a suffix, a platform or a customer list carry the markup and which do not, in a form you can join to your own data, filter down, and compare against a later copy to see what moved.