Most og:description tags are not written — they are copied. We pulled the Open Graph tags from 30 live pages and 25 of them carried a description identical, character for character, to the ordinary meta description. That shortcut stays harmless until a platform clips the tail mid-sentence. Below is what the tags actually contained, and the length range that survives every card.
Short answer: aim for 120–160 characters and write it as a card body, not as a search snippet. The measured median across those 30 pages was 137 characters.
What og:description Actually Does
The Open Graph protocol lists og:description as optional. It sits under "optional metadata" with the instruction "a one to two sentence description of your object." The four required properties are og:title, og:type, og:image, and og:url. Nothing in the spec forces a description onto the page.
Platforms do not share the spec's priorities. Of the 30 pages we could read, 28 shipped an og:description. Two did not: Wikipedia's own Open Graph protocol article, which carries og:title and og:image but no description, and theguardian.com's international front page, whose server HTML held no og: block at all — only a 62-character meta description.
Where the tag lands depends on the app. It becomes the body text under the headline in a Facebook card, the preview line in WhatsApp, the embed body in Discord, the second line of an X card. Slack reads it, Telegram reads it, LinkedIn reads it. When the tag is missing, that slot stays empty and the card simply gets shorter.
Real talk: almost nobody looks at their own link preview before shipping a page. The first time most teams see their og:description rendered is in a customer screenshot. Paste a URL into the free OG checker and you read exactly what the scraper reads — cheaper than guessing.
How Long og:description Should Be (Measured on 30 Pages)
We ran 39 URLs through the ogimgen checker and got readable page HTML back for 30 of them. Nine were served bot walls — nytimes.com, reuters.com, etsy.com, reddit.com, imdb.com, booking.com, uber.com, stackoverflow.com and producthunt.com — and were dropped from the sample. Here is how the surviving descriptions were distributed by length:
| Length | Pages | How it renders |
|---|---|---|
| Under 70 | 6 | Single short line; reads like a tagline |
| 70–120 | 6 | One full line on most cards |
| 121–160 | 9 | Two lines on Facebook; one complete sentence |
| 161–200 | 4 | Runs up to X's documented ceiling |
| Over 200 | 3 | Past the X limit; tail never renders there |
The spread inside that table is wide. Shortest string in the sample: 21 characters, on overreacted.io ("A blog by Dan Abramov"). Longest: 260 characters on react.dev. Median 137, mean 130.
X publishes the hard number. Its card markup reference caps twitter:description at 200 characters, and og:description fills that same field whenever twitter:description is absent. Three measured pages cross the line: BBC News at 249 characters, The Verge at 221, react.dev at 260. Whatever sentence runs past 200 characters in those descriptions is not doing work on X.
The short end deserves a look too. Vercel ships 45 characters, Next.js docs 37, Linear 64, Cloudflare 68. Short is not automatically a fault — a tight tagline can beat three lines of filler, and those pages know their own value proposition. What you want to avoid is landing at 21 characters because nobody filled the field and a CMS default leaked through.
og:description vs Meta Description: When One String Breaks
25 of the 28 pages with an og:description reused their meta description verbatim. Efficient, and mostly fine. The two tags do different jobs though, and the jobs collide in predictable places.
A meta description feeds a search snippet. Google rewrites it when it likes a different passage, truncates around the 155–160 character mark on desktop, and matches it to whatever was searched. An og:description is frozen: a scraper stores the string and replays it in every card, with no rewriting and no second chance to match context. Copying survives when the sentence works in both slots. It breaks when the meta description was built for keywords — the card then reads like a comma-separated list of nouns.
Twitter descriptions showed the same pattern. 23 of the 28 pages set twitter:description, and every one of them repeated the og:description word for word. So the familiar advice to hand-write a separate X description is largely unearned. Split them only when X genuinely needs a different hook, otherwise let the Open Graph fallback do its job.
One measured description closed with an emoji — joshwcomeau.com, 64 characters, heart at the end. Emoji pass through UTF-8 meta tags without drama, but the same character mangles the moment a template escapes the string twice. Confirm what the scraper sees instead of trusting your editor's preview.
og:description Fix Checklist
Seven checks, all of them cheap:
- Land between 120 and 160 characters. That range covers the measured median (137) and stays clear of X's 200-character ceiling.
- Write one sentence, with one concrete detail. A number, a product name, a promise. The measured median page described what it does, not how it feels.
- Do not restate og:title. They render stacked; duplication wastes the only line of body copy you get.
- Keep it under 200 if X matters. Past that, characters are dropped.
- Skip keyword stuffing. No card ever matched a query. Keyword-heavy meta descriptions are the main reason a copied string reads badly.
- Check the decoded value, not the source. Entities, quotes overlapping the attribute, and template double-escaping all survive your editor and die in the scraper.
- Keep all three strings in view. meta description, og:description, twitter:description — decide once whether they match.
What the Sample Says in One Line
Only 2 of the 30 readable pages omitted og:description, which means the tag is effectively universal even though the protocol calls it optional. The failures are not missing tags. They are length: 6 pages too thin to say anything, 3 pages long enough that X drops the end. Twenty-five pages copied their meta description, and that is a legitimately good default — as long as the source sentence was written for humans in the first place.
Pull your current string with the free OG checker to see the raw description, its exact length, and the fallbacks your platforms will actually use. Then build a matching 1200×630 card at ogimgen.com so the words and the image land together. Related reading: og:title covers the headline that sits above this text, og:type explains the required tag that decides how your card is treated, and the full meta tag list maps every property a scraper looks for.