什么是元标签？定义与指南

元标签是提供网页元数据的 HTML 元素。

Analytics and Measurement

A meta tag is an HTML <meta> element that gives browsers, search engines and other software information about a page, such as its description, its character encoding or how it may be indexed. It lives in the page code, and visitors never see it.

Google states that it ignores any meta tag it doesn't support, and the keywords tag that some sites still fill in has no effect on its indexing or ranking at all.

That makes a meta tag less like a command and more like a request. Whether it does anything depends on who reads it. A browser reads the viewport tag, Googlebot reads robots rules, Bingbot reads some of the same rules differently, and a social platform reads Open Graph properties when someone shares your link. This entry covers which tags matter, what each reader promises to do with them, and where the system quietly stops working.

What a Meta Tag Is, and What It Isn't

Every meta tag uses one of a few attributes. According to MDN's reference for the meta element, a name attribute supplies document-level metadata, http-equiv acts as a pragma directive that mimics an HTTP header, and charset declares the character encoding of the document. Most search-related tags use the name and content pair, as in name="description" or name="robots".

Social sharing tags work a little differently. They use a property attribute with values such as og:title, og:type, og:image and og:url, which the Open Graph protocol site defines as the basic metadata that turns a page into a rich object in a social graph. Search engines and social platforms don't necessarily read the same tags.

The Tags a Browser Cares About

Some meta tags never touch rankings. They exist for the browser. The charset tag, or its older http-equiv Content-Type form, defines the character set, and Google recommends UTF-8 where possible. The viewport tag controls how a page scales on phones, and Google reads its presence as a mobile-friendly signal. Neither one is a ranking lever. Both are basic hygiene.

Where Title and Canonical Fit

SEO tools often lump the page title and the canonical link in with meta tags. Technically, they're separate. MDN describes the meta element as metadata that can't be expressed by the other head elements, such as title, link, script and style, which is why the page title has its own <title> element and a canonical URL is declared with a link element. The distinction sounds pedantic. It isn't. Each element has its own documentation, its own rules and its own ways to break.

Which Meta Tags Google Supports, and Which It Ignores

Google publishes its list of supported meta tags, and it's shorter than you might expect. The rule attached to it is simple: you can use other meta tags if they're useful to your site, but Google ignores the ones it doesn't support.

Main Tags Google Reads

description: a short summary of the page that Google may use as the search snippet in some situations.
robots and googlebot: crawling and indexing rules. The robots name applies to all search engines, while googlebot is specific to Google.
google-site-verification: proves ownership of a site for Search Console.
charset and viewport: encoding and mobile rendering, covered above.

Tags Google Ignores

The keywords tag is the famous one. In Google's 2009 post on meta keywords, the company explained that it had stopped using the tag years earlier because it was so often abused, and its current documentation still lists it as unsupported, with no effect on indexing or ranking. Two robots values do nothing at Google either. Google no longer uses noarchive, because the cached link it once controlled no longer exists, and it says it doesn't use nocache at all.

The same logic covers community-made tags such as noai. Google doesn't list them, so Google Search ignores them. A formal vocabulary for AI usage preferences is being drafted at the IETF, but as of October 2026, the IETF AI preferences draft is still an Internet-Draft, dated September 2026, not a published standard.

Meta Description, Snippets, and the Ranking Myth

The meta description is the meta tag people argue about most. It doesn't move rankings, by Google's own account in that same 2009 post: it sometimes uses the description as snippet text, but it doesn't use the description tag in ranking. What the description can influence is how a result looks, and even that is conditional.

Why Google Shows a Different Description

Google's snippet documentation says snippets are created primarily from the page content itself. The meta description gets used when Google thinks it describes the page more accurately than text taken from the page. Snippets are also matched to the search, so the same page can show different snippets for different queries. A rewritten snippet isn't automatically a technical fault. Often it's the system working as designed.

Length: Truncation, Not a Limit

Google states there's no limit on how long a meta description can be. Google shortens the snippet when the screen requires it. So a character count is a display question, not a Google requirement. For large database-driven sites, Google calls programmatic descriptions appropriate and encourages them, as long as they're human-readable and varied, while identical descriptions repeated across many pages don't help anyone. For the craft of writing them, see our Meta Description guide, because this entry stays focused on how the tag behaves.

Robots Meta Tags: Indexing and Preview Controls

Robots rules are where meta tags carry real consequences. They decide whether a page can appear in results at all and how much of it can be previewed. Get them wrong and traffic disappears quietly.

noindex, nosnippet, max-snippet and data-nosnippet

The robots meta tag specification lists the rules Google accepts. These are the ones that come up most:

noindex keeps the page, media or resource out of search results.
nofollow tells Google not to follow the links on the page.
nosnippet removes the text snippet and video preview. At Google it applies across web search, Images, Discover, AI Overviews and AI Mode, and it also stops the content from being used as direct input for AI Overviews and AI Mode.
max-snippet:[number] caps the snippet length in characters and limits how much content can feed AI Overviews and AI Mode, except where a publisher has separately granted permission, for example through a license agreement.
data-nosnippet is an attribute, not a meta tag. It works on span, div and section elements to keep specific passages out of snippets.

When Rules Conflict

Plugins, themes and hand-coded templates can all write robots rules, so pages sometimes carry two that disagree. Google resolves this one way. The more restrictive rule wins, so a page with both max-snippet:50 and nosnippet is treated as nosnippet. It pays to look for strays, because one nosnippet left behind by an old plugin setting can silently override a snippet length you tuned on purpose.

Robots Meta vs robots.txt vs X-Robots-Tag

These three get confused constantly. The confusion is expensive. Robots meta tags and X-Robots-Tag headers are only discovered when Google crawls the URL, so if robots.txt blocks crawling, Google never sees the noindex rule, and Google's block-indexing guide names that block as a reason a page can keep appearing in results. Google also notes that it respects robots meta tags placed in the body, even though the head is the documented location. For PDFs, images and video files, which can't carry an HTML tag, the X-Robots-Tag HTTP header does the same job.

Meta Tags in AI Search: Old Tags, New Jobs

Google didn't create a new meta tag for AI answers. Instead, existing tags picked up new duties, and Google added a separate switch outside the HTML.

Google: Preview Controls Became AI Controls

Google's guide to generative AI features sets two technical conditions. A page must be indexed and eligible to show a snippet, and the site must be included in the Search generative AI setting in Search Console. That's why nosnippet now reaches further than it used to. In practice, using nosnippet to stay out of AI Overviews also strips your ordinary snippet from web search, Images and Discover.

The Search Console Setting, as of October 2026

The newer route is the Search generative AI control in Search Console, and Google says it rolled the control out to all websites worldwide as of August 31, 2026. Excluding a site removes its links and content from AI Overviews, AI Mode and generative AI features in Discover, and Google says the setting isn't used as a ranking or inclusion signal elsewhere in Search. It works per Search Console property, and a child property, such as a URL-prefix property covering one folder, can override its parent. Google also says the control doesn't govern AI training.

The regulatory backdrop is still moving. On June 3, 2026, the UK Competition and Markets Authority imposed a publisher conduct requirement on Google, which, according to the CMA's own announcement, gives publishers effective tools to prevent their content being used to power AI features in search, such as AI Overviews, with nine months for Google to implement the changes. Implementation details and deadlines may still change, so recheck the CMA announcement before relying on them. If AI visibility is a business question for you rather than a technical one, it belongs with generative engine optimization services, and our AI citation rate report offers further research on AI citations.

No Special AI Meta Tag for Google

Google's guidance is blunt. You don't need machine-readable AI files, special markup or Markdown to appear in its generative AI features, because Google Search doesn't use them, and structured data isn't required either. That statement covers Google only. It says nothing about how other AI systems behave.

Google and Microsoft Read the Same Tag Differently

Two companies, one tag, opposite outcomes. That's the clearest proof that a meta tag is a request rather than a command.

Google: noarchive Went Quiet

Google retired its cached-page link, so its robots documentation now files noarchive under historical rules that Google Search no longer uses. Adding it today changes nothing about how a page appears in Google.

Microsoft Bing: noarchive Took a New Job

Microsoft went the other way. In a September 2023 Bing Webmaster post signed by Fabrice Canel of Microsoft Bing, the company explained that content tagged NOARCHIVE would not be included in Bing Chat answers or linked from them, while content tagged NOCACHE may still be included. When a page carries both, Bing treats it as NOCACHE. Bing also said content with either tag still appears in its regular search results. As of October 2026, the 2023 post is the official source used here for this mechanism.

So the same tag that does nothing at Google can remove you from Bing's AI answers. The conflict logic differs too. Google picks the stricter rule, while Bing resolves this particular pair toward the looser NOCACHE behavior.

When Meta Tags Backfire or Don't Apply

Meta tags fail in predictable ways. Most failures look fine in a CMS settings screen. They only surface in search results later.

Blocking a page in robots.txt to reinforce noindex. The block hides the noindex rule, so the URL can still be listed through links from other pages.
Using nosnippet as an AI opt-out. It works, but it also removes your standard snippet across Google surfaces. The Search Console setting separates the two.
Filling in meta keywords. Google gives the tag no weight at all.
Adding noarchive just in case. Google ignores it, while Bing keeps the content out of Bing Chat answers, which may not be what you wanted.
Injecting or changing meta tags with JavaScript. Google recommends avoiding this where possible and testing thoroughly when you can't.
Stamping one description across hundreds of pages. Google says identical descriptions aren't helpful when individual pages appear in results.
Trusting noai tags to keep content from Google. Google ignores tags it doesn't support.

A Three-Question Check

Before trusting any tag, ask three things. Was it delivered, meaning does the HTML Google receives actually contain it? Google requires the head section to be valid HTML for its metadata to be machine readable. Could the crawler see it, or did robots.txt stop the fetch? And which reader honors it: Google, Bing, a browser or a social platform? The URL Inspection tool in Search Console shows the HTML Googlebot received, which answers the first two questions for Google. For site-wide audits, this check belongs in the technical layer of search engine optimization services.

Meta Tag FAQs

Do meta tags affect Google rankings?

Mostly no. Google doesn't use meta keywords or the meta description for ranking. Robots rules are different, because noindex decides whether a page can appear in results at all.

How long should a meta description be?

Google sets no length limit. It truncates snippets to fit the device, so write a complete, accurate summary rather than chasing a character count.

Why does Google show a different description than mine?

Google builds snippets primarily from page content and matches them to the query. It uses your meta description when it judges that text to be the more accurate summary.

Is the title tag a meta tag?

Not technically. The page title uses its own <title> element, which sits next to meta tags in the head but follows separate rules.

What's the difference between a robots meta tag and robots.txt?

Robots.txt controls whether a URL can be crawled. A robots meta tag gives indexing and preview rules once the page is crawled. If robots.txt blocks the URL, the meta tag is never read.

Do I need a special meta tag for AI Overviews?

No. Google says no special AI markup is needed. To appear, a page must be indexed and snippet-eligible, and the site must be included in the Search generative AI setting.

How can I check which meta tags Google sees?

Use the URL Inspection tool in Search Console. It shows the HTML Googlebot received, including any robots rules.