Open graph tags on 26 homepages: 19 carry all four required properties
Open graph tags are nearly universal and nearly complete. On 26 homepages read 2026-08-20, 19 carry all four required properties, 25 carry the optional og:description, and og:url is the one that gets dropped.
FIELD TEST · 2026-08-20 · 30 sites · single crawl
Sample and method: the same 30 sites used since 2026-08-15 for the robots.txt, sitemap and homepage structure surveys. One homepage request each with a browser user agent, 2026-08-20, from an exit in Japan, with the <head> parsed out of the raw HTML. Same crawl as two other posts in this batch.
Open graph tags are nearly universal on this panel and nearly complete. Twenty-six sites delivered a real homepage. Nineteen of them carry all four properties the protocol calls required; 25 carry og:description, which the protocol calls optional. The one that gets dropped is og:url, missing on five.
How we measured it
Fixed before the crawl, unchanged after.
- Request
https://<domain>/once, follow redirects, browser user agent, and keep the response body. - Read every
<meta>in the returned HTML and bucket it bypropertyorname:og:*,twitter:*, and plaindescription. - Count a site only if the body it returned is an actual homepage, not a challenge page.
That last rule removed two sites that answered with HTTP 200. www.canva.com returned 3,038 bytes titled "Client Challenge" and www.reddit.com returned an 8,393-byte shell; neither carries a single meta tag. Two more never got that far: stackoverflow.com and medium.com both answered 403. So 30 requested, 28 answered 200, 26 usable.
Which open graph tags the protocol requires, and who ships them
The protocol names four. On 26 homepages read 2026-08-20, all four are present together on 19 of them, and every single one of the four is present on more than 20.
| Property | Required | Sites |
|---|---|---|
og:title | yes | 25 / 26 |
og:description | no | 25 / 26 |
og:image | yes | 24 / 26 |
og:type | yes | 23 / 26 |
og:url | yes | 21 / 26 |
og:site_name | no | 14 / 26 |
og:image:alt | no | 3 / 26 |
The protocol's own wording, read 2026-08-20 at ogp.me: "The four required properties for every page are: og:title … og:type … og:image … og:url". og:description appears further down under a heading that begins "The following properties are optional for any object and are generally recommended".
The property everybody ships is the one the specification never asked for.
Who is missing what
Seven sites fall short of the required set, and only one of them is missing more than a single property.
| Site | Missing |
|---|---|
| news.ycombinator.com | all four — no og: tags at all |
| linear.app | og:type, og:url |
| developer.mozilla.org | og:type |
| www.bbc.com | og:image |
| discord.com | og:url |
| webflow.com | og:url |
| www.netlify.com | og:url |
Hacker News is the clean outlier: no open graph tags, no twitter tags, a 34,324-byte page and a plain <title> reading "Hacker News". Everything a machine gets about that page, it gets from the body.
One more thing worth naming, because it is a bug rather than an omission. slack.com sets og:image to /dev-cdn/v/marketing/img/homepage/… — a path, not a URL. The protocol asks for "An image URL". Every other site in the panel that sets the property sets it absolute.
og:description is a copy of the meta description, every time
On all 25 homepages that carry both, the two strings are byte-identical. Not similar. The same string, twice, in the same document.
That is worth knowing before you spend an afternoon writing a separate social description. In this panel nobody did, including the sites whose whole business is publishing. If the two are always going to match, the interesting question moves upstream to the description itself, which we measured separately in the meta description length survey.
When og:title differs from the title, it is shorter
Twenty-five sites carry og:title. On 21 of them it matches <title> exactly. On the other four it is the same sentence with the brand suffix cut off.
| Site | og:title | <title> |
|---|---|---|
| arstechnica.com | Ars Technica | Ars Technica - Serving the Technologist since 1998… |
| www.wikipedia.org | Wikipedia, the free encyclopedia | Wikipedia |
| github.com | …GitHub keeps you ahead. | …GitHub keeps you ahead. · GitHub |
| www.wired.com | WIRED - The Latest in Technology… | WIRED - The Latest in Technology… | WIRED |
Wikipedia runs it the other way: the og:title is the longer of the two. Three of four strip the brand, one adds it. That is not a convention, that is four independent decisions. We measured how long those title strings run in the meta title length survey.
Twitter tags are still there, and they mostly say one thing
Twenty-two of 26 carry twitter:card. Twenty of those set it to summary_large_image and two to summary. Below that, twitter:description appears 20 times, twitter:title 20, twitter:image 19, twitter:site 15.
The pattern is a full open graph block plus a thin twitter block that overrides the card shape and repeats the same text. Nobody in this panel writes genuinely different copy for the second block.
og:site_name is the entity signal, and half the panel skips it
Fourteen of 26 declare og:site_name. It is the only open graph property that names the publisher rather than the page, which makes it the one worth caring about if you want an answer engine to attach a page to the right organisation. The stronger version of that declaration lives in structured data, and how to write it is in entity SEO for AI search.
What this crawl cannot tell you
We did not test whether any AI engine reads these tags. Open graph is a social preview protocol; using it as a proxy for how a model describes your page is an assumption, and this measurement does not support it either way.
It is also one request per site on one day. Canva alone has produced a 3 KB shell and a 436 KB homepage on two consecutive days in earlier runs, so a single snapshot of a head section is exactly as fragile as the page it came from.
What to check on your own homepage
Four things, in the order that finds problems fastest.
- Confirm
og:urlexists and is absolute. It is the property this panel skips most, and it is the one that tells a consumer which address the page claims to be. - Confirm
og:imageis a fullhttps://URL. Slack's relative path is the failure mode to copy from, in the sense of not copying it. - Do not write a separate social description before checking whether anything renders it differently. In this panel the two strings are identical 25 times out of 25.
- Do not treat a page as machine-readable because it returned 200. Two sites here returned 200 with no head worth reading.
Checking every page on a site rather than one homepage is what QueryWin is being built to do.
Common questions
How did you measure this?
One HTTPS request per homepage, browser user agent, redirects followed, 2026-08-20, from a single exit in Japan. Meta tags parsed from the raw HTML with no JavaScript executed, so anything injected client-side is not counted.
Which open graph tags are actually required?
Four: og:title, og:type, og:image, og:url. That list is from the protocol document at ogp.me, read 2026-08-20, and it has not changed in years.
Does Google use open graph tags for ranking?
Google has never documented open graph as a ranking input, and this crawl tested nothing about ranking. What we can say is narrower: 25 of 26 sites in this panel put the same description string in both places, so the two are not independent signals in practice.
Is og:description worth writing if it is optional?
On the evidence here it is written more often than three of the four required properties, which tells you what publishers believe. Whether that belief is correct is a different measurement, and we have not run it.
