Every few months someone sends me a page that looks perfect in the browser and terrible everywhere else — a naked gray rectangle on X, the wrong headline on Google, a Facebook preview showing the site's cookie banner. Nine times out of ten the fix is a handful of meta tags.
Meta tags are the labels in your page's <head> that tell search engines and social networks what your page is. Most guides list forty of them. In 2026, you genuinely need about a dozen. Here's what each one actually does, according to what the platforms currently read — and which famous ones are zombies you can delete. (If you'd rather skip the theory, our free Meta Tags Generator builds the whole block for you with live Google, Facebook, and X previews.)
The two that carry your search result
<title> is still the single most important line of metadata on the page. Google usually shows the first 50–60 characters — it truncates by pixel width, not letters, so a title full of Ws runs out sooner than one full of Is. Put the meaningful words first; leave the brand name for the end.
<meta name="description"> doesn't affect ranking — Google confirmed that years ago and it's still true — but it's your ad copy. When it's missing or weak, Google writes its own snippet from whatever page text it likes, which is how cookie banners end up in search results. Aim for one or two sentences, roughly 150–160 characters, that make a human want to click.
The traffic cops
<link rel="canonical"> tells search engines which URL is the "real" one when the same content is reachable at several addresses (?utm_source=, trailing slashes, http vs https). Without it, your page competes against its own duplicates. One per page, always absolute.
<meta name="robots"> only matters when you don't want the default. index, follow is what crawlers assume anyway; noindex is the one that has teeth. The classic footgun is shipping noindex from staging to production — if a page mysteriously vanishes from Google, check this tag first.
<meta name="viewport"> (width=device-width, initial-scale=1) isn't SEO decoration — without it, phones render your page as a zoomed-out desktop site, and Google's mobile-first indexing notices.
Open Graph: how your link looks everywhere else
Facebook invented Open Graph, but in 2026 it's the lingua franca — Facebook, LinkedIn, Slack, Discord, WhatsApp, iMessage, and most chat apps all read it. Six tags do the real work:
og:titleandog:description— the card's headline and blurb. They can be friendlier than your search title; this is social, not a results page.og:image— the big one, literally. 1200 × 630 pixels is still the sweet spot in 2026. It must be a real, publicly reachable URL — not a base64 data URI (crawlers reject those outright, which is a bug we once fixed in our own tool) and not localhost.og:image:width/og:image:height— optional but smart: platforms can render the card on the first share instead of waiting until after they've downloaded the image.og:url— the canonical address of the page, so shares consolidate to one URL.og:type—websitefor most pages,articlefor posts. Deep taxonomy beyond that is largely ignored now.
Add og:image:alt for screen-reader users and og:site_name for the small print above the headline. That's the whole set.
X cards: two tags, then fallbacks
X (Twitter) reads its own twitter:* tags first, then quietly falls back to Open Graph for anything missing. In practice you need exactly two:
twitter:card—summary_large_imagefor the big picture card,summaryfor the small square one. If the page has a decent 1200 × 630 image, use the large card; it gets dramatically more engagement.twitter:site— your @handle, shown as attribution.
twitter:title, twitter:description, and twitter:image are only worth setting if you want them different from Open Graph. Otherwise, let the fallback do its job.
The zombie graveyard 🪦
Tags you can delete today, guilt-free:
<meta name="keywords">— dead since 2009. Google has said flatly that it isn't a ranking signal; if anything, a spammy keywords tag is a small red flag. Yet it's still the most-pasted meta tag on the internet.revisit-after— never was a real thing. No major crawler has ever honored it.rating,distribution,expires— relics of 1990s directories.generator— free reconnaissance for attackers ("this site runs WordPress 5.2"), zero benefit.
Deleting these won't boost your ranking, but every line of <head> you don't ship is one you don't have to maintain.
The mistakes that actually bite
After years of looking at broken previews, it's almost always one of these:
- The og:image URL is relative (
/images/card.png). Crawlers need the fullhttps://address. - The image is behind a login or a robots block, so the crawler sees a 403 and shows gray nothing.
- A typo'd protocol like
https:/example.com(one slash) — browsers forgive it, crawlers don't. - Stale cache — Facebook and X cache cards aggressively. After fixing your tags, run the URL through the platform's sharing debugger to force a refresh, or you'll conclude your fix "didn't work."
- No image dimensions, so the first share of every new post renders as a plain text card.
Skip the memorizing
You could keep all of this in your head — or you can fill in four fields and watch live previews of exactly how your page will look on Google, Facebook, and X, then copy one finished block of HTML. That's what our free Meta Tags Generator does, including the pixel-width truncation, the image checks, and the handle normalization. It runs entirely in your browser, and like everything on this site, there's no sign-up and no catch.