Main area
Agency
on-page-seo

Meta Tags Done Right — Title, Description, Canonical, Robots and Friends

Redaktion ·

Why meta tags still matter

Two half-truths haunt the meta-tag conversation. One: “Title and description are everything — the rest is legacy.” Two: “Google ignores meta tags anyway, so why bother.” Both are wrong. Title and description remain the most direct click levers because they shape the SERP snippet. At the same time, canonical, robots, hreflang and Open Graph solve technical questions that, without these tags, either go unanswered or get resolved by the search engine in ways you won’t like — and that’s where painful duplicate-content, indexation and sharing problems start.

This article weaves the relevant meta tags into a single picture. We move from the classic must-haves (title, description) through the technical control tags (canonical, robots, hreflang) to the social-sharing tags (Open Graph, Twitter Card) — and look at what AI Overviews and answer engines have changed.

The basics: what meta tags actually are

Meta tags are HTML elements inside <head>. They give machines — search engines, social crawlers, browsers — information about the page that isn’t part of the human-facing body. Some are display hints (description as a snippet suggestion), some are control directives (noindex), some resolve ambiguity (canonical for duplicates, hreflang for multilingual variants).

The crucial distinction is hint vs. directive. Google obeys robots directives (noindex, nofollow) — they are followed. A description is only a suggestion: Google can rewrite the snippet when another passage on the page matches the query better. This asymmetry is the source of much confusion — once you internalize it, you stop asking “why didn’t Google use my description.”

Two families: SEO meta and social meta

Roughly, meta tags fall into two families:

  • SEO meta — talks to search engines: title, description, canonical, robots, hreflang, X-Robots-Tag (HTTP header).
  • Social meta — talks to social platforms and messenger crawlers: Open Graph (Facebook, LinkedIn, WhatsApp, Slack), Twitter Card (X/Twitter).

The families overlap (a good title helps everywhere) but need different tags and different lengths. If you only maintain SEO meta, you lose clicks when links are shared. If you only maintain social meta, you have a pretty LinkedIn preview but a weak Google snippet.

Title and description — the two biggest click levers

The title tag and meta description are the only meta tags users see directly — in the SERP, in browser tabs, in bookmarks, in shared links. Everything else is machine communication. That’s why these two deserve different care than the rest.

Title — the page’s headline

The title sits inside <title> and is the first bold link in the SERP. Practical rules of thumb:

  • Length: 50–60 characters visible (pixel limit ~580 px desktop, ~330 px mobile). Anything longer gets truncated, usually with ….
  • Keyword early: Put the primary keyword in the first 30 characters where it has the strongest click and ranking signal.
  • Brand at the end: Separate with | or –. For brand queries Google often flips the order for you anyway.
  • Uniqueness: One distinct title per page. Duplicate titles across a domain signal duplicate content and dilute click-through.

Google rewrites titles roughly 60 % of the time today. Common triggers: titles that are too long, keyword stuffing, generic brand-stacks (“Homepage | Brand”), or a better H1 that matches the query. Carefully written titles get rewritten less; sloppy ones hand the decision over to Google.

Description — the click amplifier

The meta description lives in <meta name="description" content="…"> and provides the snippet under the title. It is not a classical ranking factor (Google has said so consistently since 2009) but a massive click factor.

  • Length: 150–160 characters. Mobile shows less (~120), desktop up to ~160.
  • Active voice: Verbs, concrete promise, USP. No filler like “Learn more about X on our site.”
  • Include the keyword: Not for ranking, but because Google bolds matching keywords in the snippet — measurably lifts click rate.
  • Uniqueness: Each page its own description. For product pages with thin data, leaving it blank often beats copying — Google then auto-generates from the page.

When the description is empty, Google builds a snippet from matching passages. On well-structured content this is often better than a stilted description. On flat product data or repetitive copy, a deliberately written description is clearly superior.

Canonical — the anti-duplicate insurance

The canonical tag is one of the most powerful and most-misused meta tags. It sits in <link rel="canonical" href="…"> and tells search engines: “This URL is the canonical version of this content — when you find the same content under multiple URLs, consolidate all signals to this one.” Concrete use cases:

  • Parameter URLs: /product?sort=price → canonical points to /product.
  • Filter variants: Faceted navigation in shops generates dozens of URL variants of the same content.
  • HTTPS / trailing slash: One version per piece of content, the others canonicalize there.
  • Print versions, AMP variants: Canonical to the main version.
  • Cross-domain duplicates: Syndicated articles on partner sites can canonicalize back to the original.

Self-referencing canonical

Best practice: every indexable page carries a canonical pointing to itself. It feels redundant but is insurance — crawlers, tracking parameters and CMS quirks add URL variants without anyone noticing. A self-referencing canonical resolves that case immediately.

Common canonical mistakes

In practice, four mistakes show up over and over:

  • Wrong target: Canonical points to a page with different content → Google ignores the signal or follows it blindly. Both are bad.
  • Canonical to a 404 or redirect: Breaks the signal entirely.
  • Canonical to a noindex page: Contradictory signals. Google decides for itself, usually not in your favor.
  • Canonical with a relative URL: Allowed but risky with CDN/subdomain setups. Use absolute URLs.

Robots meta and X-Robots-Tag — the indexation switches

Where canonical says “this is the right version,” the robots meta tag says “should/shouldn’t be in the index.” It lives either in the HTML (<meta name="robots" content="…">) or as an HTTP header (X-Robots-Tag, required for PDFs, images and other non-HTML documents).

Most important values:

  • index, follow — default. Index the page, follow links.
  • noindex — keep the page out of the index. Typical for filter pages, internal search results, thank-you pages, employee logins.
  • nofollow — don’t pass link equity through links on this page. Rarely sensible at page level — nofollow is usually a link attribute, not a page attribute.
  • noarchive — no cache snapshot. Rarely used.
  • nosnippet / max-snippet:0 — no snippet displayed. Also affects AI Overviews.

The most common self-inflicted wound: noindex on production

By far the most common meta-robots production bug: a staging site runs with noindex on every page, goes live, nobody removes the tag. Weeks later the team wonders why nothing ranks. Pre-launch routine: spot-check 10 important URLs for index, follow — fastest via curl -I or browser devtools.

Hreflang — the language and region switch

Hreflang is the meta tag for multilingual and multi-regional sites. It tells Google: “This page exists in language X too, here’s the URL.” Format: <link rel="alternate" hreflang="en-us" href="…">.

Three rules that account for 90 % of hreflang errors:

  • Reciprocal: If DE points to EN, EN must point back to DE. One-way hreflang entries are ignored.
  • Self-reference: Each language variant also references itself — otherwise the consistency check breaks.
  • x-default: A default variant for unsupported languages/regions. Often the global English version or a language-picker page.

Hreflang is not a ranking booster — it’s a serving logic. Google still ranks by the classic factors, but with multiple matching languages it picks the right variant for the user’s region and language.

Open Graph and Twitter Card — the sharing aesthetic

When someone pastes a URL into WhatsApp, LinkedIn, Slack or Discord, those crawlers pull the preview — image, title, description — from Open Graph tags. Without them, shared links appear as a bare URL with no image or title, which crushes click rate.

Minimum set for any shareable page:

<meta property="og:title" content="…">
<meta property="og:description" content="…">
<meta property="og:image" content="https://…/og.jpg">
<meta property="og:url" content="https://…">
<meta property="og:type" content="article">
<meta name="twitter:card" content="summary_large_image">

Image requirements

The og:image is the single biggest click lever in sharing. Recommended specs:

  • Format: JPG or PNG, not SVG (sharing crawlers often don’t render SVG).
  • Size: 1200 × 630 pixels (ratio 1.91:1) for Facebook, LinkedIn, Twitter Large Card. Minimum 600 × 315.
  • File size: under 5 MB, ideally under 300 KB for fast crawling.
  • Content: Readable even at small sizes — keep text on the image to roughly 20 % of the area.

LinkedIn and WhatsApp cache Open Graph data for days. If a new og:image doesn’t show up, the LinkedIn Post Inspector or WhatsApp’s debugging tool can invalidate the cache.

Structured data and sitemap — the supporting cast

Closely related to classic meta tags but technically distinct: structured data (JSON-LD in <head> or body) and the sitemap (XML file, not in the HTML head).

Structured data is what enables rich results in the SERP — star ratings on products, FAQ boxes, breadcrumb paths, how-to steps. For AI Overviews, schema.org markup is becoming increasingly important: answer engines prefer to cite clearly structured sources marked up with Article, FAQPage or HowTo schema.

The XML sitemap isn’t a meta-tag question, but it belongs to the on-page tag discipline: it lists every indexable URL, gets submitted in Search Console, and speeds up the crawling of new and changed pages.

Reference table — what goes where

| Tag / element | Location | Required? | Purpose | |---|---|---|---| | <title> | <head> | Yes | SERP headline, browser tab | | meta description | <head> | Recommended | SERP snippet | | link rel="canonical" | <head> | Recommended (self-ref) | Duplicate control | | meta robots | <head> | When needed | Index/follow control | | X-Robots-Tag | HTTP header | For non-HTML files | Robots meta for PDFs/images | | link rel="alternate" hreflang | <head> | For multilingual sites | Language/region variant | | Open Graph (og:*) | <head> | Recommended | Social sharing preview | | Twitter Card (twitter:*) | <head> | Recommended | X/Twitter sharing | | Structured data (JSON-LD) | <head> or body | Recommended | Rich results, AI answers | | <meta charset="utf-8"> | <head> | Required | Encoding | | <meta name="viewport"> | <head> | Required (mobile) | Responsive rendering |

Three scenarios played through

Scenario A — blog article on an agency site

The article should rank well and get shared well. Minimum set:

  • Title: “Search Intent Explained — Keywords Are Just the Input | boostN” (~58 characters, keyword early, brand at the end)
  • Description: “How keyword research becomes real SERP traction: intent classes, long-tail clusters, format choice — explained step by step.” (~127 chars)
  • Canonical: self-referencing
  • Robots: index, follow (default)
  • Hreflang: DE → EN, EN → DE, x-default → EN
  • Open Graph: slightly shorter title, custom 1200×630 image with the article title as text overlay
  • Structured data: Article schema with author, datePublished, publisher

Scenario B — shop filter page

/running-shoes?color=blue&size=42 — should not be indexed because dozens of filter URLs would otherwise compete:

  • Title: can be generic, won’t rank anyway
  • Robots: noindex, follow — don’t index but keep crawling links so product detail pages remain reachable
  • Canonical: to /running-shoes (the canonical filter-less category)
  • Hreflang: not applicable — a filter page isn’t a real language variant

Scenario C — multilingual landing page for lead generation

/en/services/seo/conversion/ — should be discoverable in DE and EN, runs on paid social:

  • Title and description: conversion-driven, clear promise, brand at the end
  • Canonical: self-referenced
  • Robots: index, follow
  • Hreflang: DE ↔ EN reciprocal
  • Open Graph: purpose-built sharing image that works in LinkedIn feeds (light background, large type, short hook)
  • Structured data: Service schema with rating (when real reviews exist) for rich results

FAQ

Why doesn't Google use my description?
Because it didn't match the user's query as well as another passage on the page. Google replaces the snippet dynamically when the SERP query targets a different part of the content. The description is a suggestion, not a directive.
How long can a title be today?
50–60 characters visible. Longer titles get rewritten or truncated — both cost clicks. If sub-branding or long brand names are necessary, put them at the end, where truncation hurts least.
Do I still need Open Graph now that Twitter is X?
Yes. Open Graph is used by LinkedIn, Facebook, WhatsApp, Slack, Discord, Telegram and more — not just Facebook. Twitter Card is X/Twitter only and can often be skipped when og:image and og:title are set (X partially reads Open Graph as fallback).
What happens to meta tags in AI Overviews?
Title and description still shape the source attribution that answer engines show. Structured data is becoming more important because answer engines prefer structured sources. nosnippet / noindex blocks AI answers too — check before applying.
Are keywords meta tags still relevant?
No. <meta name="keywords"> has been officially ignored by Google since 2009. Bing doesn't use it meaningfully either. Time spent on keywords meta is time wasted.

Conclusion

Meta tags aren’t a historical artifact — they’re the only layer where you tell machines directly how to interpret a page. Title and description remain the biggest click levers because they shape the SERP. Canonical, robots and hreflang resolve technical ambiguities that, without these tags, devolve into duplicate-content and indexation problems. Open Graph decides whether a shared URL gets clicks in social feeds or dies as a naked text link.

The most practical recommendation: bake meta-tag setup into the launch checklist of every new page — not as “we’ll do it later.” Title, description, canonical, robots, hreflang (if multilingual), Open Graph with an image. Six fields, set in ten minutes, paying off in clicks and indexation safety for months.

If you want to go deeper: read up on content structure as the sister topic — headings, semantic HTML and content hierarchy form the other half of on-page SEO.

Themenuebersicht

On-Page SEOMeta-Tags