A meta tag is an HTML element that sits inside the <head> of a page and describes that page to browsers, search engines and other software, without appearing in the text the user reads.
Of all the meta tags you can write, Google Search uses about ten. It discards the rest: it neither rewards nor penalises them. The serious problem on real websites is not choosing the wrong tag. It is writing it in a part of the page where Google has already stopped reading, which happens on 10.1% of desktop pages.
What a meta tag looks like
It is a void element: it opens and never closes. It comes in three forms, depending on the attribute it carries.
<meta name="..." content="...">is a name/value pair:namesays what the information is,contentholds it.<meta http-equiv="..." content="...">simulates an HTTP header, as if the server had sent it in the response.<meta charset="utf-8">declares the document's character encoding.
On the last form the HTML specification is stricter than most guides suggest. The value must match utf-8, the only valid encoding for an HTML5 document, and the declaration must sit entirely within the first 1,024 bytes of the document: past that point the browser has already read it some other way. This is documented by MDN Web Docs, which follows the WHATWG HTML standard.
A point that is rarely made: the page title is not a meta tag. It is the <title> element, which has a closing tag and text inside it, and does not use the name and content attributes. Meta title is SEO jargon, not HTML. The rules for writing it do not change, but if someone tells you it is missing and you look for a <meta> in the source, you are looking at the wrong line.
The meta tags Google actually uses
This is the full list of meta tags supported by Google Search, taken from the Google Search Central documentation, last updated on 8 March 2026.
| Meta tag | Syntax | What it does | Ranking factor |
|---|---|---|---|
| description | <meta name="description" content="..."> | Can become the snippet under the title in the results | No |
| robots / googlebot | <meta name="robots" content="noindex"> | Controls crawling and indexing. The googlebot version applies to Google only | No, but it decides whether the page appears |
| viewport | <meta name="viewport" content="width=device-width, initial-scale=1"> | Tells the browser how to scale the page on mobile. Signals to Google that the page is designed for mobile | Indirect |
| charset | <meta charset="utf-8"> | Declares the character encoding | No, but if it is wrong special characters break |
| notranslate | <meta name="googlebot" content="notranslate"> | Asks Google not to offer a translated version of the page in the results | No |
| google-site-verification | <meta name="google-site-verification" content="..."> | Verifies site ownership in Search Console | No |
| refresh | <meta http-equiv="refresh" content="0;url=..."> | Client-side redirect. Google recommends replacing it with a server-side 301 | No |
| rating | <meta name="rating" content="adult"> | Flags adult content so that SafeSearch can filter it | No |
| nopagereadaloud | <meta name="google" content="nopagereadaloud"> | Stops Google's text-to-speech services from reading the page aloud | No |
None of these moves a page up the rankings. Two of them, though, change something that matters: the meta description decides what the user reads before clicking, and the meta robots tag decides whether there is anything to click at all.
The ones Google ignores, and why you still find them everywhere
The well-known case is <meta name="keywords">. Google stated that it does not use it for ranking on 21 September 2009, in a post signed by Matt Cutts, and its position has not changed since: today it appears in the documentation under "unsupported tags and attributes". If you are wondering whether to remove it, we covered that in meta keywords.
The same list contains three more, less talked about:
- the
langattribute: Google detects language from the text content, not from the annotation in the code. That does not make it useless, since screen readers and browsers use it, but it is not a signal for Search. To tell Google which language version to show to whom, you need hreflang. rel="next"andrel="prev": Google no longer uses them for indexing paginated pages.nositelinkssearchbox: the search box it used to switch off no longer exists, so the rule has no effect.
Open Graph tags (og:title, og:image and the rest) are a separate matter. Google Search does not read them; Facebook, LinkedIn and WhatsApp do when someone pastes the link. You should write them, but for a reason that has nothing to do with rankings.
One piece of advice about this whole category keeps circulating: keep the <head> lean, because useless tags confuse the search engine. That is not how it works. Google processes the meta tags it recognises and discards the rest, with no consequences. The damage comes from something else.
The point where meta tags stop working
The HTML standard allows only eight elements inside <head>: title, meta, link, script, style, base, noscript and template. If Google meets any other element, for example a <div>, it assumes the <head> has ended there and stops reading the elements that follow.
The consequence is harsh: any description, canonical, hreflang or meta robots below the invalid element effectively ends up in the <body> and no longer does anything. The tag is written correctly, it is in the source, you can see it with Ctrl+U, and for Google it does not exist.
This is not a textbook case. The HTTP Archive Web Almanac 2025, which analyses a global crawl, found invalid HTML inside the <head> on 10.1% of desktop pages and 10.3% of mobile pages. The elements that break it most often are <img>, <div> and <a>, and they are almost never written by hand: they come from a plugin, a tag manager or a badly inserted third-party script.
You check it with the URL Inspection tool in Search Console, looking at the rendered HTML rather than the served source. If a meta tag sits below an invalid element, either move it above or remove the element. This kind of check belongs to technical SEO, not content work, even though the symptom looks like a content problem.
When two meta tags say different things
A page can carry both <meta name="robots">, aimed at all search engines, and <meta name="googlebot">, aimed at Google only. When the directives point in opposite directions, Google applies the more restrictive one. The example in the documentation: if a page declares both max-snippet:50 and nosnippet, nosnippet wins and the snippet disappears entirely.
The default values are index and follow, so declaring them achieves little. In the Web Almanac crawl index appears on 69% of desktop pages and follow on 64%, while noindex, the only directive that actually changes the page's status, is on just 3.5%. We have written about what it does and when it fails in noindex.
A warning worth repeating: meta robots works on a page the crawler is able to read. If you block the URL in robots.txt, Googlebot does not download the page, never sees the noindex, and the URL can stay in the results without a description.
How widespread they are, in numbers
Again from the Web Almanac 2025 global crawl:
| Element | Desktop pages | Mobile pages |
|---|---|---|
<title> | 98.62% | 98.54% |
| meta description | 67.7% | 67.2% |
| meta viewport | 93.1% | 95.2% |
noindex directive | 3.5% | 2.4% |
Invalid HTML in the <head> | 10.1% | 10.3% |
The meta description has the most eventful history: it was on 71% of pages in 2022, fell to 66.7% in 2024 and rose again to 67.7% in 2025. A third of the web still does not write one.
The most interesting figure is a different one, though, and it concerns length. The median <title> is 77 characters on desktop and 79 on mobile, well beyond the "safe" 50-60 character range repeated in every guide. It is not that the web got it wrong en masse: many sites have simply stopped optimising for truncation, since Google rewrites the displayed title anyway when it sees fit.
Where to write them if your site runs on a CMS
On WordPress, Shopify or Wix you almost never touch the <head> HTML by hand. Yoast SEO and similar plugins expose two fields per page, title and meta description, plus a switch for noindex. Google's own documentation points to this route rather than editing the code directly.
Two things to check once, and then you can forget about them:
- That there are not two SEO plugins active at the same time. The result is two
<meta name="description">tags on the same page, and Google picks one without telling you which. - That the theme does not print its own description on top of the plugin's. It happens with themes that include their own SEO module, and the symptom is the same: two tags where you need one.
On JavaScript, Google's documentation is openly cautious: you can add or change meta tags with JavaScript, but it is better to avoid it where possible, and if you must, test the implementation. This applies above all to sites built on frameworks that assemble the <head> client-side.
Meta tag FAQs
No. The page title is the <title> element, which has a closing tag and does not use the name and content attributes. Meta title is a common expression among SEOs, but in the code there is no <meta name="title"> that Google reads.
None of the meta tags supported by Google is a direct ranking factor. The meta description affects how many people click the result, meta robots decides whether the page appears, and viewport signals that the page is designed for mobile. These are real effects, but they work through something other than the page's score.
You do not have to: Google has ignored it since 2009 and does not penalise it. Removing it has one practical benefit, which is that you stop showing competitors the list of keywords you are working on.
In most cases three are enough: charset, viewport and description. Meta robots is added only where the page needs to be kept out of the results, and google-site-verification only on the page used for verification in Search Console.
Google uses the description text when it considers it a better answer to the query than an extract taken from the page; otherwise it generates the snippet from the content. It is not an error to fix in itself: it is worth acting only if the description is generic or has nothing to do with the query.
The way it is usually told, writing meta tags sounds like a copywriting job: finding the right words within a character limit. On the sites we work on, the time is almost always lost elsewhere, and it is always the same three things: a plugin that doubles the description, a script that breaks the <head>, a noindex left over from launch day. Looking at the rendered HTML of ten pages pays off more than rewriting the first description, and it takes half an hour. If you would rather have someone check it for you, it is part of what we do in our SEO services.
