What the metadata check tells you
Every website describes itself in a few places: a title and description in its head, a language on its html element, icons for tabs and home screens, a web app manifest, structured data naming the organisation behind it, and links to its social profiles. Directories, link previews, browsers and search engines all read these signals. The metadata check reads them the same way and puts them on one page, with the site's logo picked from the best source available.
What we check
- The basics. The page title, the meta description, the canonical address, the declared language and
the site name, taken from
og:site_nameor else from the Organization or WebSite name in the site's structured data. - Social profiles. Links to profiles on networks such as LinkedIn, X, GitHub, YouTube and Instagram,
taken from the structured data
sameAslist and from links in the page header, navigation and footer. Links markedrel="me"count too, which is how Mastodon profiles are usually confirmed. We list at most 20, and never read email addresses or phone numbers. - Icons. The favicon, declared icons with their sizes, the Apple touch icon and the icons in the web app manifest. We fetch up to six distinct candidates, check each one loads and is a real image, and read its size from the file.
- The manifest. Its name, short name, colours and icon list, and whether it is valid JSON at all.
- The logo. The first candidate that works, in the order the FAQ below describes. When the only usable picture is the page's preview image we use it, but flag it, because a sharing image rarely makes a good logo.
How images are shown
Logos and icons are never loaded from the site in your browser. We fetch each file once on our side and make our own copies: a 64 pixel preview of every working icon, and 256 and 128 pixel PNG copies of the logo, all stripped of metadata. Vector logos keep their shape as a cleaned SVG copy; we remove scripts, event handlers, embedded HTML and links to other sites before showing it, and tell you when something had to go. Every copy travels inside the result, so a shared report shows the same images later.
Reading the result
The logo comes first, with where it came from and its size. Under it are the facts, with "Not set" where the site doesn't provide one, then the social profiles, then every icon candidate in the order we tried them. Each icon shows our small copy or what went wrong: a 404, a file that isn't an image, or an SVG that needed cleaning. The manifest section appears when the site links one.
Findings cover what's missing or broken: no favicon at all, a declared icon that doesn't load, no Apple touch icon, a logo that had to fall back to the preview image, a manifest that can't be read and an SVG with scripts. The check has no score; the findings are facts about the site's markup, with fixes and guides for the ones worth acting on.
What the check doesn't do
We read only the home page and the files it points to, with an honest user agent, and we don't run JavaScript. Icons added by scripts after the page loads aren't seen. When the home page can't be fetched the check stops with a plain message and you aren't charged. When an icon or the manifest can't be fetched because it timed out or the host asked us to slow down, the result is marked partial and says which.