Meta tagging SEO comes down to four things Google actually consumes: the title tag, the meta description, the robots meta tag, and the canonical link. Everything else in the head of your document is either for social platforms, for your own tooling, or ignored entirely. The meta keywords tag has been dead since 2009 and no amount of blog folklore revives it.
Get those four right on every indexable page and you have captured essentially all the ranking and click-through value available from your document head. Here is what each one does, the exact lengths to write to, and the mistakes that quietly cost traffic.
The title tag: your highest-leverage 60 characters
The title tag is a genuine ranking input and the strongest driver of click-through rate from the results page. Google rewrites titles it considers unhelpful — studies of large samples have put rewrite rates somewhere north of half of all queries — but a well-built title survives far more often than a padded one.
Write to under 60 characters. Desktop results truncate around 580 pixels, which lands near 60 characters for typical letter mixes; capital-heavy titles clip earlier. Put the primary keyword near the front, keep the brand at the end after a pipe or dash, and drop it entirely when the title is already tight. Skip the keyword stuffing pattern — “Blue Widgets | Buy Blue Widgets | Blue Widgets Online” reads as spam to users and gets rewritten by Google anyway.
The meta description: no ranking value, real click value
Google has said plainly that meta descriptions are not a ranking factor. They still matter, because they occupy two lines of prime results-page real estate and can lift click-through rate meaningfully on pages where the competition writes lazy copy.
Target 140–155 characters. Google truncates around 155–160 on desktop and shorter on mobile, so writing to 155 gives you room without risk. Include the target phrase once — matched terms are bolded in the snippet, which draws the eye. Lead with the outcome, not the setup: “Cut your audit time from a day to twenty minutes” beats “We are a leading provider of audit services.” And write one per page. Duplicate descriptions across a template are a wasted slot, and Google will replace them with page text.
Robots meta and canonical: the tags that break sites
These two carry the least glamour and the most risk. A stray noindex pushed from staging can remove a category tree from search inside a week, and nobody notices until traffic charts flatten.
- noindex — keeps a page out of the index. Correct for thin filters, internal search results, and thank-you pages. Never combine with a
disallowin robots.txt; if the crawler cannot fetch the page it never sees the directive. - nofollow on the meta tag — rarely useful in 2026. Handle link-level control at the link.
- max-snippet, max-image-preview, max-video-preview — control how much of your page can appear in results and AI-generated summaries. Publishers use these; most sites should leave them at defaults.
- rel=canonical — declares the preferred URL for duplicate or near-duplicate content. It is a hint, not a command. Self-reference every canonical page, and make sure the canonical URL is absolute, indexable, and returns a 200.
Open Graph and Twitter cards
Open Graph tags do nothing for Google rankings and everything for how your page looks when someone pastes the URL into Slack, LinkedIn, or a group chat. That sharing behavior drives real referral traffic and the occasional link, so treat these as cheap insurance rather than SEO.
Ship four at minimum: og:title, og:description, og:image, and og:url. Use a 1200×630 image so it renders sharp on every platform. Twitter falls back to Open Graph when its own tags are absent, so a single twitter:card value of summary_large_image is usually all the extra markup you need.
Tags you can stop writing
A surprising amount of head markup in production templates does nothing at all. Removing it is a small speed win and a large clarity win.
- meta keywords — ignored by Google since 2009, and a free look at your keyword strategy for anyone reading source.
- meta revisit-after — never supported by any major engine.
- meta author, meta copyright — no search effect. Keep them if internal systems read them.
- Generic meta robots index, follow — that is already the default. It only adds bytes.
- Language meta tags — use the
langattribute on<html>and hreflang link elements instead.
Auditing meta tags at scale
On a twenty-page site you can check by hand. Past a few hundred URLs you need a crawl, because the failures cluster in templates and only show up in aggregate. Run a full-site crawl and sort for six conditions:
- Missing title tags
- Duplicate titles across more than one URL
- Titles over 60 characters
- Missing or duplicate meta descriptions
- Unintended
noindexdirectives on indexable templates - Canonicals pointing to redirects, 404s, or noindexed URLs
What separates a useful audit from a useless one is evidence. A count of “142 duplicate titles” is a number; a list showing the actual title text, the exact URLs, and the H1 on each page is a work queue. SEO Rocket runs an instant quick scan of roughly 25 pages and deep crawls verified past 900 pages, and reports the real strings behind every issue so you can fix the template rather than the symptom.
Writing titles and descriptions that earn clicks
Once the mechanics are right, the copy is what separates a 3% click-through rate from a 7% one at the same position. A few patterns hold up across categories.
- Lead with specificity. Numbers, years, and constraints outperform adjectives. “7 Rank Trackers Compared on Price and Accuracy” beats “The Best Rank Trackers.”
- Match the query’s wording in the title where it reads naturally, because matched terms bold in the snippet and the eye goes there first.
- Do not promise what the page does not deliver. Inflated titles buy a click and lose the visitor in eight seconds, and Google notices the pattern eventually.
- Use the description to handle the objection. If your competitors all charge more, say the price. If the reader’s worry is setup time, say how long it takes.
- Skip the brand suffix on informational pages where every character is doing work. Keep it on commercial and homepage-adjacent pages where recognition helps.
Test rather than debate. Change the title on one page, wait four weeks, and compare click-through rate at a comparable average position. Anything under a couple of hundred impressions in the window will not give you a readable result.
A practical workflow that scales
Meta tagging fails at scale because it is written by hand for the first fifty pages and by a bad template for the next five thousand. Fix that with three rules.
First, set template defaults that are decent rather than clever — product name plus category plus brand, built from fields you already have. Second, hand-write titles and descriptions for the pages that earn real traffic; the top 50 URLs by impressions in Search Console usually cover most of the opportunity. Third, re-crawl monthly and diff against last month, because meta regressions arrive through deploys, not through neglect.
Use Search Console as ground truth for whether a rewrite worked. Pull the page’s impressions and click-through rate for the four weeks before and after the change, and ignore anything smaller than a couple of hundred impressions — the variance will eat the signal. If you want that crawl-plus-fix loop, meta validation on generated content, and rank tracking sitting in the same workspace, SEO Rocket handles it at one flat price. Either way, the tags themselves are a solved problem: write four of them well, delete the rest, and spend the saved time on the pages.
