Paste a page's HTML and see how its link will share — and why it might not. Open Graph problems are almost all silent: when a tag is wrong the platform shows no preview and reports nothing, so the only clue is a link that looks bad. The commonest cause by far is a relative og:image, which a crawler has no page context to resolve and therefore ignores completely, while the tag sits there in the markup looking perfectly fine. That is checked here, along with the mixed-content case, and the resolved absolute URL is offered where the page address is known.
| Tag | If missing or wrong |
|---|---|
| og:image | No preview at all — must be absolute |
| og:title | Falls back to the page title |
| og:description | Falls back to the meta description |
| og:url | Share counts split across entries |
| og:type | Defaults to website |
| twitter:card | Falls back to Open Graph |
Nine times out of ten the og:image is a relative URL. A crawler reading your tags has no page to resolve it against, so it ignores it and shares without a preview, reporting nothing. It must be absolute, including https.
Caching. Platforms cache what they scraped and will keep serving the old preview until it is refreshed. Use Facebook's Sharing Debugger or LinkedIn's Post Inspector to force a re-scrape; X refreshes on its own schedule. This tool reads your tags rather than their cache, which is exactly how you tell the two apart.
Usually not. Twitter falls back to Open Graph for title, description and image, so a separate set is only needed to override something — a different image for that platform, or a large summary card. Duplicating every tag adds maintenance for no gain.
1200×630 is the safe answer everywhere. Facebook wants at least 600×315 and recommends 1200×630; that size also crops acceptably to the squarer and wider shapes other platforms use.
More than it looks. Platforms use it to deduplicate shares, so a missing or inconsistent value splits the share count for one page across several entries. It should be the canonical address of the page.
Because a browser cannot read another origin's HTML. Anything offering to check a URL from the browser is sending it to a server, which then has a record of every URL you tested. Pasting from view-source keeps it in the page.
A close reconstruction rather than a live render. Each platform crops and styles differently, and shows its own cached version. The tag checking is the reliable part.
Open Graph problems are almost all silent, which is what makes a checker useful: when a tag is wrong the platform shows no preview and reports nothing, so the only signal is a link that shares badly. The single most common cause is a relative `og:image`. A crawler fetching your tags has no page context to resolve a relative path against, so it ignores the value entirely — and because the tag is plainly present in the markup, the fault is invisible to anyone reading it. The same applies to an image served over plain HTTP from an HTTPS page, which several platforms refuse as mixed content. Both are checked, with the resolved absolute URL offered where the page's own address is known. Beyond the image: `og:title` and `og:description` fall back to the page title and meta description on most platforms, so their absence is recoverable rather than fatal and the tool says so instead of marking everything red. `og:url` matters more than it looks, because it is what platforms use to deduplicate shares — a missing or inconsistent value splits the share count for one page across several entries. Twitter reads Open Graph as a fallback, so a separate `twitter:` set is only needed to override something, which is worth knowing before duplicating every tag. The markup is pasted rather than fetched for the same reason as the meta analyser: a browser cannot read another origin's HTML, so a tool offering to check a URL is proxying it through a server that then sees every URL you test. Parsing uses the browser's own HTML parser, and reads `property` as well as `name` — an extractor looking only at `name` reports every Open Graph tag as missing, since the specification uses `property`. The preview is a reconstruction rather than a live fetch, so it shows what the tags say rather than what a platform's cache currently holds, which is a distinction worth understanding when a share still looks wrong after a fix.