When someone pastes your website's URL into WhatsApp, LinkedIn or Slack and a card appears with
an image, a title and a description, that card is not generated by the browser or decided by the
search engine. It is generated by the social platform itself, at the moment the link is pasted (or
shortly after), by reading a specific set of <meta> tags in your page's
<head>. That mechanism has a name of its own: Open Graph.
If you'd rather fix this yourself right now, try our free tool Open Graph Preview (social media).
If you need professional help with this, at 3CWeb we offer a Technical SEO service.
What the Open Graph protocol is
Open Graph is a protocol created by Facebook in 2010 so that any web page could become an
"object" within the social graph: something with a title, image, description and type, that could
be shared, liked or commented on consistently. Technically it consists of a group of
og:* properties declared as RDFa meta tags in the HTML
<head>, shaped like this:
<meta property="og:title" content="..." />
Although Facebook created it, today it is read by practically every platform that generates a link preview: LinkedIn, WhatsApp, Slack, Discord, Telegram, Pinterest and many more. Twitter/X uses its own system (Twitter Cards, explained further below) but can inherit from Open Graph when it does not find its own tags. In practice, Open Graph has become the de facto standard for link previews across the social web, even though it never went through a standards body such as the W3C.
How it differs from SEO meta tags
Here lies the most common confusion, and it is worth clarifying it from the start because it
shapes everything else: Open Graph is not SEO. These are two separate tagging
systems, aimed at different audiences, that coexist in the same <head> but do
not replace one another.
<title>and<meta name="description">are for search engines: Google decides, with a certain amount of freedom, what text to show as the title and description on its results page, and uses the content of these tags as the main candidate (although it may rewrite them if it believes there is more relevant text elsewhere on the page).<link rel="canonical">is for telling crawlers which URL is the "official" version of a page when several URLs exist with the same or very similar content (with tracking parameters, with and without a trailing slash, on http and https, and so on). It affects which URL gets indexed and how ranking signals are consolidated across duplicates.- The
og:*tags are for social networks and messaging services: they decide what title, description and image are shown in the preview card when the link is shared. Google does not useog:titleorog:descriptionto decide your position in the search engine.
You can (and usually should) use similar wording across both, but they are not interchangeable: the SEO title is usually optimised for a keyword and for not exceeding a length limit in the search engine; the Open Graph title can be somewhat more eye-catching or editorial, because its job is to earn the click in a social feed, not to rank in a SERP.
The essential Open Graph tags
These are the ones any page should carry, at minimum, so that its social preview looks right on any platform:
og:title
The title shown on the card. It can match the SEO <title> or be a shorter,
more direct variant; many platforms truncate it past a certain length (roughly 60-90 characters
depending on the network), so the main idea should sit in the first few words.
og:description
The description under the title. As with og:title, the SEO meta description text
can be reused, but some platforms cut it more aggressively (WhatsApp and Slack tend to show
noticeably less text than LinkedIn), so the first sentence needs to stand on its own.
og:image (with og:image:width and og:image:height)
The preview image. It is, by far, the element that most influences whether someone clicks a
link shared on social media. Explicitly declaring the image's real width and height with
og:image:width and og:image:height matters for a very concrete reason:
some platforms render the card before fully downloading the image, and use those declared
dimensions to reserve the right space and decide the crop. If you do not declare them, or if they
do not match the file's actual dimensions, the typical result is a badly cropped, stretched, or
oddly centred image, especially in the "large image" format used by Twitter/X and Facebook.
og:url
The canonical URL of the content being shared. This matters because, if the same piece of
content is reachable from several URLs (for example with campaign tracking parameters), always
declaring the same og:url makes Facebook and other platforms accumulate likes and
comments under a single social graph object, instead of splitting them across URLs that a human
sees as "the same page" but that a social platform treats as separate objects.
og:type
The type of object being shared. The most common values are website for general
pages (such as the homepage or a landing page) and article for editorial content such
as a blog post. The article type also enables additional, specific properties, such
as article:published_time, article:author or article:section,
which some platforms use to enrich the card or to group content by date or author.
og:site_name
The name of the site or brand, independent of the specific page's title. Several platforms show it as a discreet label next to the card (for example, under the image or next to the domain), so it is clear which website the content comes from even if the title does not mention it.
og:locale
The content's language and region, in language_TERRITORY format (for example
en_US or es_ES). Facebook uses it, among other things, to decide how to
format dates or auxiliary card text according to regional settings. On a multilingual site, each
language version should declare its own correct og:locale, consistent with the
page's other language signals (such as the HTML lang attribute or
hreflang tags).
Twitter Cards and their relationship with Open Graph
Twitter (now X) developed its own tagging system before fully aligning with Open Graph, and it
still keeps its own twitter:* namespace. The main ones are:
twitter:card: the card type. The most common value for most pages and blog posts issummary_large_image, which shows a large image across the full width of the card; there is alsosummary, with a small thumbnail image next to the text, closer to the classic preview style used by other networks.twitter:titletwitter:descriptiontwitter:image
The important thing to understand here is the inheritance relationship: if a page does not
declare its own twitter:* tags but does have og:* tags, Twitter/X can
automatically fall back to the values of og:title, og:description and
og:image to build the card. That is why many websites simply declare
twitter:card with the value summary_large_image and let the rest be
filled in from Open Graph, instead of manually duplicating the same text and image under two
separate namespaces. This is a legitimate and very common simplification, although explicitly
declaring twitter:title and twitter:description gives more control if
you want different wording specifically for that platform.
Recommended image sizes (and what happens if you get them wrong)
For og:image, the de facto reference is an aspect ratio of
1.91:1, with a technical minimum of 600 x 314 pixels and a
recommended size of around 1200 x 630 pixels, which offers the best sharpness on
high-density screens without producing an unnecessarily heavy file. For Twitter/X with
summary_large_image, a similar ratio is recommended, also close to 1.91:1 or 2:1,
with 1200 x 630 as a safe choice that works well across both systems at once.
What happens in practice if these requirements are not met:
- If the image is smaller than the platform's required minimum, some networks simply discard it and show the card without an image, or fall back to a generic image if a backup one is defined on the site.
- If the ratio is very different from 1.91:1 (for example a square or very vertical image), the automatic crop can cut off exactly the important part of the image, such as a face or some text, because the crop centres on the middle without knowing what is actually in the photo.
- If a page has no
og:imagetag at all, the platform usually tries to guess a reasonable image from the HTML itself (for example the first large image in the content) and, if it cannot find anything it considers valid, the card is shown without an image, with only a title and description.
How to actually debug this
The most common mistake when checking these tags is looking only at the source code (whether the tags are written correctly) rather than at how the platform actually interprets them, because there is a factor at play that does not depend on your website at all: each social network's cache.
The reference tools for checking this properly are:
- The Facebook Sharing Debugger (part of Meta's developer tools), which shows
exactly which
og:*tags Facebook has read on your URL and lets you force a fresh crawl ("Scrape Again"). - The Twitter Card validator, which serves the same purpose for the
twitter:*namespace (its public availability has changed over time following the transition to X, so it is worth checking whether it is still accessible or whether an equivalent tool exists at the time you use it). - LinkedIn's and other platforms' validators, with the same logic: they show what the platform currently has cached and, in some cases, let you invalidate that cache.
And here lies the real cause behind a very common problem: "I changed the page's image but
Facebook/WhatsApp still shows the old one." This is almost never a fault in the website itself.
These platforms cache the result of the first crawl of a URL for quite a long time
(it can be days or weeks) so they do not have to re-download and re-process the page every time
someone shares the same link. If you change og:image and do not force a fresh crawl
with the corresponding debugger, the network will keep showing the cached version until it decides
to refresh it on its own, which can take much longer than seems reasonable. The fix is not to
change something else on the server, but to open that specific platform's debugging tool and
explicitly ask it to re-crawl the URL.
Example of a tag block in HTML
A complete block, for a blog post, would look like this in the
<head>:
<!-- SEO (independent of Open Graph) -->
<title>Search-engine-optimised title | Site name</title>
<meta name="description" content="Description written for Google's SERP.">
<link rel="canonical" href="https://yourdomain.com/blog/my-article">
<!-- Open Graph -->
<meta property="og:type" content="article">
<meta property="og:title" content="Title written for sharing on social media">
<meta property="og:description" content="Short, direct description for the social card.">
<meta property="og:image" content="https://yourdomain.com/img/my-article-og.jpg">
<meta property="og:image:width" content="1200">
<meta property="og:image:height" content="630">
<meta property="og:url" content="https://yourdomain.com/blog/my-article">
<meta property="og:site_name" content="Site name">
<meta property="og:locale" content="en_US">
<!-- Twitter Cards -->
<meta name="twitter:card" content="summary_large_image">
<meta name="twitter:title" content="Title written for sharing on social media">
<meta name="twitter:description" content="Short, direct description for the social card.">
<meta name="twitter:image" content="https://yourdomain.com/img/my-article-og.jpg">
The most common misunderstanding: thinking this "helps SEO"
It is very common to come across the idea that filling in Open Graph tags properly "improves
SEO" or "helps ranking". That is not the case, at least not directly. Google does not use
og:title, og:description or og:image as a ranking factor:
they are tags meant for a different consumer (social media crawlers, not Googlebot) and are not
part of the signals the search algorithm evaluates to decide a page's position.
What is true, and why it is still worth getting right, is that a social card that looks good (with a clear image, a title that invites a click, and a description that is not cut off halfway through a sentence) earns more clicks when the link is shared on social media or messaging apps. That better click-through rate when sharing is not a Google ranking signal, but it is a real behavioural signal: more clicks mean more visits, more time on the site, more chances that someone links to the content from elsewhere, and that, indirectly and through several steps, can end up benefiting the content's visibility. The key distinction is causality: Open Graph does not move rankings directly, it improves the sharing experience, and that experience can produce positive side effects.
Frequently asked questions
Do I need to duplicate the title and description, or can I reuse the SEO ones?
You can reuse the same text from <title> and
meta name="description"> in og:title and og:description
with no technical issue; in fact that is what most pages do. The only reason to separate them is
if you want a different tone for social media (more eye-catching or shorter) compared with the
text written for Google's results page, where the available space and the text's purpose are
different.
Do Open Graph tags affect SEO?
Not directly. Google does not use them as a ranking factor. Where they can have an influence is indirectly, through behaviour: an attractive social card generates more clicks when the link is shared, which means more traffic and more visibility, but that is a side effect of people's behaviour, not a direct use of those tags by the search algorithm.
How do I stop a social network from still showing an old cached image?
It is not fixed by changing more things on the server: you need to explicitly force a fresh crawl in that platform's debugging tool (for example, the "Scrape Again" button in the Facebook Sharing Debugger, or the equivalent in the Twitter Card validator). Until that fresh crawl is requested, the platform will keep serving the version it has cached, no matter how long ago the real image on your server already changed.
Do I need a different og:type for a blog article than for the homepage?
Yes, that is the recommended practice: the homepage and general site pages usually use
og:type with the value website, while blog posts and editorial content
use article. The article type also enables additional, specific
properties, such as the publish date or the author, which some platforms may display or use to
enrich the card.
What happens if I don't set og:image:width and og:image:height?
The platform may have to download the full image before it can render the card at the correct size, which sometimes causes a different crop or scaling than you expected, especially if the file's actual dimensions do not match the recommended 1.91:1 ratio. Explicitly declaring the real width and height helps the platform reserve the correct space from the start and reduces the chance of an unexpected crop.