Open Graph preview tool: see how your link looks when it gets shared

Who this page is for: anyone whose tags are already live and who wants to know what each network does with them before sending the link to a list, a group or an ad. If you still have to write the tags, start at the meta tag generator, which builds the whole head block ready to paste. If your question is about the Google results page rather than social networks, the route is the SERP snippet preview, which measures the title cut in pixels. The three tools look at different parts of the same head: one writes it, one measures search, and this one measures sharing.

The same link becomes six different cards

There is a comfortable illusion that Open Graph is a standard, and that a standard means identical behaviour. It does not. The tags are the same, but each platform decides three things on its own: which tag it reads, where it cuts the text and how it crops the image. The result is that the same link can show up in full on Slack, with a truncated title on WhatsApp and with no description at all on LinkedIn, without a single error anywhere in your code.

Three examples that come up every day. LinkedIn does not show og:description on the feed card: what is left is image, title and domain, and a title that depended on the second line to make sense arrives half finished. X, with twitter:card set to summary_large_image, draws the title over the image itself and also shows no description. WhatsApp does accept a description, but it has the shortest text budget of them all plus a practical file weight limit that makes the image vanish with no warning.

None of those three things shows up in your editor, in your CMS or in an HTML validator. They only show up when someone pastes the link. This tool moves that moment earlier.

How to use it

1. If your tags are already live, open paste the head here, paste the meta tag block and hit read the tags. The fields fill themselves in. If they are not live yet, type straight into the fields.
2. Enter the dimensions and the weight of the image. It is the input almost no preview asks for and it is exactly what decides whether the card gets an image at all. If you prefer, pick the file under preview with the real image and all three fields fill themselves in.
3. Walk through the platform selector. The card is redrawn on every switch, with the format, the truncation and the crop of that network.
4. Read the same tags across every platform table. It answers in one glance where your text survives in full and where it gets cut, with how many characters disappear in each place.
5. Fix whatever shows up under what to fix before sharing and watch the score climb. The score measures the tag set; truncation is a separate reading, field by field.

How it works: the three decisions every platform makes

Everything the tool does comes from four simple rules applied with each network's reference numbers:

The fourth rule is the one that surprises people. Social networks do not stretch images: they crop from the centre. A 1200 by 630 piece of art, which is 1.90:1, fits a 1.91:1 frame almost exactly and loses 0.3% of its area. That same art, forced into a square, loses 47.5% of its width, and everything on the sides disappears. Since you do not get to choose which format your link lands in, the practical rule is to design for both: essential information inside the central square, breathing room at the edges.

Character counting uses perceived characters, not code points: an emoji with a skin tone counts as one, which is how the platform cuts and how a human counts. And the cut prefers to release the last whole word rather than split it in half, which is the behaviour observed on real cards.

Worked example (reproduces the default output)

The preloaded example is the link of a running shoes guide, with a 1200 by 630 pixel image at 240 KB, on the WhatsApp platform. Checking what the tool shows, number by number:

• Title: 72 characters, against the WhatsApp limit of 65. The card shows up to "on a" and stops there: 10 characters disappear, taking the word treadmill with them.
• Description: 135 characters, against the limit of 100. The line ends at "price" and 39 characters disappear, including the cushioning table, which was half the argument.
• Image: ratio 1.90:1, a 0.3% crop in the 1.91:1 frame of the large card. In the diagram the green outline covers almost the whole art and the dashed square cuts 47.5% of it.
• Tag score: 92, band A, with two caveats in the list: title and description are both over the tightest limit across platforms.

Now walk through the platform selector without touching anything else and watch the same link change behaviour. On LinkedIn the title shows in full, because the limit there is 100 characters, but the description disappears entirely. On X the title loses the same 10 characters, and the description does not appear either: the large card is an image with the title over it. On Facebook, Telegram and Slack everything fits in full. Six platforms, three different results, one single block of tags.

Run two more tests to see the two failures nobody catches in time. Change the file weight from 240 to 420 KB and go back to WhatsApp: the card loses the image and becomes plain text, with a note saying the file is heavier than the platform will fetch. Then put 240 back and change the dimensions to 800 by 800: the image turns square, the crop in the large card frame jumps to 47.6% and the score drops from 92 to 83.

Reference measurements, platform by platform

These are the values the tool uses, checked in September 2026. No platform publishes them as a contract, and all of them change the card without notice, so treat the limits as orders of magnitude. What is not an estimate is the tag chain: that is the documented reading rule of each network and it is what explains most unexpected previews.

• WhatsApp: large card when the image is at least 300 by 200 pixels and the file stays under roughly 300 KB. Title around 65 characters, description around 100. It is the tightest of the six, in both senses.
• LinkedIn: large card at 1.91:1, title up to around 100 characters, no description in the feed. Recommends 1200 by 627 pixels.
• X (Twitter): the format comes from the twitter:card tag. On summary_large_image, a large image with the title over it and no description, title around 70 characters. On summary, a square thumbnail with title and description beside it.
• Facebook: large card, title around 88 characters on desktop and far fewer on mobile, description around 155, domain written in uppercase in the footer.
• Telegram: card with the site name on top, title around 90 characters and description around 200. It is the most generous with text after Slack.
• Slack: shows the site name, title around 80 characters and description up to roughly 300. The image only renders large when it is wide, from about 500 pixels up.

How to read each number

The social tag score measures the set, not one platform's card: the tags exist, they point at a file that can actually load, and the text fits the tightest budget. You can score well and still see truncation on one network, which is why the verdict line under the card speaks per platform rather than for the set.

The characters that disappear answer the question the percentage cannot: what exactly is being lost. Ten characters at the end of a title are usually one word; thirty nine at the end of a description are usually the reason someone would click. If what disappears is irrelevant, so is the cut. If what disappears is the promise, the fix is almost always moving the information to the front, not deleting words.

The image area cut only becomes a problem past roughly 20%. Below that the crop is eating margin; above it, it starts eating content. The square crop number is the worst case and works as a safety ceiling: if your art survives the square, it survives every card.

The tag that won is the most underrated row in the report. When it shows the title came from the document title tag rather than og:title, it means the card text is the text you wrote for Google, brand name at the end and all. That is usually the explanation for the odd card nobody can account for.

Known limits

The biggest limit is cache. Every platform stores the card for days after the first read, so a link that already circulated keeps showing the old version even after the tags change. Facebook, LinkedIn and X have official tools to force a refetch; WhatsApp and Telegram do not, and there the trick is versioning the image URL.

Each platform's measurements are references, not contracts. The same link renders slightly different cards in the app and in the browser, on phone and on desktop, and the networks adjust those layouts with no announcement. The difference between 88 and 92 characters changes no writing decision; the difference between showing and not showing the description does, which is why the tool separates those two levels of information.

The tool also does not fetch your URL or download your image: it works with what you enter. That is the price of having no server in the middle, and it is a deliberate trade, because it means the tags of an unpublished page never touch a machine that is not yours. To check the whole head after fixing the card, the meta tag generator returns the complete block, and the Article schema generator handles structured data, which is a separate matter.

From a good looking card to a tested click

A card that fits is the floor, not the goal. Once the link shows in full everywhere, the next question is which version makes more people click and what happens to the ones who do:

• If the destination is a landing page, run it through the landing page grader before fighting for more clicks to a page that does not convert.
• Before deciding whether one page version beats another, check how much traffic the test would need in the sample size calculator.
• With the test live, get the verdict from the statistical significance calculator instead of calling it on the first good day.
• To find out whether the site is ready to be cited by AI systems, and not just shared, use the GEO readiness checker.

Frequently asked questions

Why does my link show up with no image on WhatsApp?

There are four causes, in that order of frequency. First is file weight: WhatsApp fetches the image in a short window and gives up when the file is too large, and in practice the preview starts failing intermittently somewhere above 300 KB. Second is the URL: the tag has to point to an absolute https address, because the platform reads the tag outside the page context and cannot resolve a relative path. Third is image size: below 300 by 200 pixels the card falls back to the square thumbnail or disappears. Fourth is cache: if the link already circulated once with the wrong tag, the stored version keeps being served for a good while. Enter your image weight and dimensions in the tool above and the card changes right away, showing which of the four is yours.

What is the ideal Open Graph image size?

1200 by 630 pixels, which gives the 1.91:1 ratio Facebook, LinkedIn and X recommend for the large card, with a file under 300 KB so it survives WhatsApp. That format passes on every platform with no meaningful crop. The detail almost nobody accounts for is the square crop: when a platform falls back to the thumbnail it crops the centre at 1:1 and throws away nearly half the width of your art. That is why the practical rule matters more than the number: keep the logo and any text inside the central square and the image works in both formats.

Does LinkedIn show my meta description on the card?

Not in the feed. The LinkedIn link card today is image, title and domain, and og:description does not appear there. That changes what you write: on LinkedIn the title has to stand on its own, because there is no second line to finish the argument. The same reasoning applies on X with the large card, which also draws only the title over the image. Keep filling in og:description, because WhatsApp, Facebook, Telegram and Slack do use it, but do not count on it to close the promise.

Do I need twitter: tags if I already have Open Graph?

You need one: twitter:card. The rest are optional, because X falls back to the Open Graph tags when it finds no twitter: version. Without twitter:card, though, it will not build the large card even with the right image, and you lose the format that takes the most room on screen. twitter:title and twitter:image only earn their place when you want different text or different art on X, which is a legitimate but rare case. The tool above shows which tag won on each platform in the report rows, precisely so you can see when the text on screen is not the text you thought you had written.

I changed the tags and the preview is still the old one. Now what?

Every platform caches the card for days, and some for weeks. The way out is forcing a refetch in each official tool: the Facebook sharing debugger, the LinkedIn post inspector and the X card validator all have a scrape again button. WhatsApp and Telegram have no public tool, and there the move is changing the image URL, adding a version to the filename, because the cache key is the address. Before you go forcing caches, check here whether the new tags actually solve the problem: forcing a refetch of a still wrong tag only swaps one bad card for another.

Does the tool fetch the tags from my URL?

No, and that is deliberate. Fetching a URL would need a server of ours in the middle, and then the text of your unpublished page would pass through it. Here everything runs in your browser: you paste the meta tag block and parsing happens on your machine, with no network call. The image, if you pick a file, is read from your local disk through a temporary browser address and is never uploaded anywhere. You can confirm it by opening the network tab while you use the tool: nothing goes out.

Is og:image:alt worth anything?

It is, for two things. First is accessibility: it is the text a screen reader announces in place of the art when someone shares the link, and without it the person only hears that an image exists. Second is diagnosis: a well written alt forces whoever produced the card to answer what is in the image, and that question usually reveals that the art communicates nothing on its own. It does not influence ranking or card size, so anyone promising that is selling something that does not exist.

How many characters does an emoji count for in og:title?

One, and that is the right count. The tool uses perceived characters, which is how a human counts and how platforms cut on screen. A common emoji, a flag or an emoji with a skin tone can occupy two, four or more code points, and a naive counter would tell you your flag eats four characters of the budget. Keep in mind, though, that an emoji takes more width than a letter: if your title is already at the edge, the emoji can push the cut earlier than the number suggests.

Keep reading

To understand how search and AI answers treat what you publish, follow up with CRO vs SEO, what changes in each, read the guide to generative engine optimization and see B2B SaaS landing page A/B testing.