nosnippet: keep a page out of AI Overviews

For one page, the tag is nosnippet, not Google-Extended. Google says a page can appear as a link in AI Overviews or AI Mode only if it is indexed and eligible to be shown in Search with a snippet. Take the snippet away and the page is out of both. The cost is that the same tag also removes the snippet from ordinary search results. For a whole site or folder, Google added a Search Console setting in 2026.

Why blocking Google-Extended does not do it

Many sites add Google-Extended to robots.txt expecting to leave AI Overviews. Google's crawler documentation says what that token controls: whether content may be used to train Gemini models and to ground answers in Gemini Apps and on Vertex AI. It then says, in so many words, that Google-Extended "does not impact a site's inclusion in Google Search".

AI Overviews and AI Mode are part of Search, so they follow the controls Search already has. Google's page on AI features names them:

To limit the information shown from your pages in Search, use nosnippet, data-nosnippet, max-snippet, or noindex controls.

Blocking Google-Extended is still a reasonable choice if you do not want your pages used for Gemini training. It just answers a different question.

For a whole site: the Search Console setting

A whole site or folder can leave Google's AI answers without tagging every page. Search Console Help says Google rolled the control out to all websites by 31 August 2026. It sits under Settings > Search generative AI and covers AI Overviews and AI Mode, among other generative AI features. Include is the default. Excluded content is not shown, linked or used for grounding (Giving a model content from the search index while it writes an answer, so the answer rests on it.)Giving a model content from the search index while it writes an answer, so the answer rests on it. there, and gets no traffic or impressions from them.

It is set per Search Console property. A URL-prefix property such as https://example.com/blog/ inherits from its closest configured parent and can override it, so a folder can opt out on its own. A single page cannot, which is why the tag still matters.

Google expects the change within one to two days, some pages later because of caching. It is not a ranking or inclusion signal for the rest of Search, and it leaves training alone:

This control doesn't affect AI training; to limit training of the models used to generate responses in Search generative AI features, use Google-Extended.

To leave Search completely, Google's help points to noindex.

The four controls, from widest to narrowest

noindex

noindex removes the page from Search altogether, AI features included. It is the widest control and the most expensive one: the page earns no search traffic at all.

<meta name="robots" content="noindex">

nosnippet

nosnippet keeps the page indexed but shows it without a text snippet. Google's specification says it applies to every form of search result, AI Overviews and AI Mode included, and that it also prevents the content from being used as a direct input for them. The page can still rank, with its title and URL and no description.

Put it in the page's head, or send the same directive as an HTTP header, which is the only way to set it on a PDF or any other file that has no head:

<meta name="robots" content="nosnippet">

X-Robots-Tag: nosnippet

max-snippet

max-snippet caps the snippet at a number of characters, and Google says the cap also limits how much of the content may be used as a direct input for AI Overviews and AI Mode. Two values are special: max-snippet:0 is the same as nosnippet, and max-snippet:-1 lets Google choose. The cap does not override permission you have granted separately, for example through structured data or a licence agreement.

<meta name="robots" content="max-snippet:120">

data-nosnippet

data-nosnippet works on part of a page instead of all of it. It goes on a span, div or section, and the text inside is left out of snippets while the rest of the page stays available.

<p>This sentence can appear in a snippet.</p>
<div data-nosnippet>
  <p>This answer stays on the page and out of snippets.</p>
</div>

It has three traps:

• It is a boolean attribute, so its value is ignored. data-nosnippet="false" still hides the text.
• The HTML has to be valid. An unclosed div with the attribute hides everything after it.
• It does not work on other elements. A p or a custom element with the attribute is ignored.

Google also warns against adding or removing it with JavaScript, because the attribute may be read before or after the page is rendered.

Bing and Copilot read other tags

Everything above is Google's. For Copilot, Bing's Webmaster Guidelines give each tag a different reach:

• noindex keeps a URL out of Bing search, Copilot experiences and grounding API results.
• noarchive keeps the content out of Copilot responses and grounding results.
• nocache limits Copilot to the URL, title and snippet.
• nosnippet and data-nosnippet stop Bing from showing captions, and Bing says they "may limit Copilot citation quality". It does not say they keep a page out of Copilot.

Bing added data-nosnippet in October 2025. Its announcement says marked text "is still indexed normally, but it will be excluded from snippets and AI summaries". Each tag also works as an X-Robots-Tag header, and Bing's list of supported tags shows how to scope one to Bing:

<meta name="bingbot" content="noarchive">

Choosing one

• You want the page out of search entirely. Use noindex, and accept that it is gone from ordinary results too.
• You want the page findable but not summarised. Use nosnippet. Expect fewer clicks from ordinary results, where the page now shows without a description.
• You want a whole site or folder out of AI Overviews and AI Mode. Use the Search Console setting.
• You want a short teaser and a cap on reuse. Use max-snippet with a small number.
• You want to protect a few passages, such as a price table or the paid part of an article. Put data-nosnippet on those blocks and leave the page alone.
• You want your pages kept out of Gemini training. Block Google-Extended in robots.txt.
• You want a page kept out of Copilot answers. Use noarchive, or noindex to leave Bing altogether.
• You want to be cited. Use none of them. For most pages that is the right setting, and the default.

How the rules combine

When rules conflict, the most restrictive one applies. A page with both max-snippet:50 and nosnippet is treated as nosnippet.

Rules for different crawlers add up. A robots tag that says nofollow and a googlebot tag that says noindex together mean noindex, nofollow for Googlebot.

The header and the meta tag count the same. A CDN, a framework or a server rule can add an X-Robots-Tag that nobody sees in the HTML, so check the response headers as well:

$ curl -sI https://example.com/pricing | grep -i x-robots-tag

Put together, a page meets Google's gates in this order:

flowchart TB
  A("Crawl allowed?") -- "No" --> X["Tags never read"]
  A -- "Yes" --> B("noindex?")
  B -- "Yes" --> Y["Out of Search"]
  B -- "No" --> C("Snippet blocked?")
  C -- "Yes" --> Z["Out of AI answers"]
  C -- "No" --> D("Site excluded?")
  D -- "Yes" --> Z
  D -- "No" --> E["Eligible, not guaranteed"]

How geo.new checks it

An audit reads the robots and googlebot meta tags in the page's source HTML and the X-Robots-Tag header, whether it applies to every crawler or is scoped to Googlebot. Two checks in AI Access and Rendering use them:

• ACC-11, Snippet and preview availability. The catalog reads "pass none or scoped data-nosnippet (noted) · warn finite max-snippet · fail nosnippet / max-snippet:0". A data-nosnippet is noted but not warned, because it often marks a cookie banner rather than the content.
• ACC-10, Search indexing directives. The catalog reads "pass none · warn conflicting declarations · fail noindex applies". The failure is printed beside the score, so a good average cannot hide it.

Both feed the Google AI Overviews line in Platform Readiness. A failure on either is a question, not a verdict: if keeping the page out of AI Overviews is what you want, the check has found exactly that. The sample report shows both rows on a fixture page, each with the evidence it read.

An audit reads the page alone, so it cannot see the Search Console setting, and it does not read tags scoped to bingbot. A noarchive in a robots tag or header is read: CORE-14 warns on it and asks you to keep it only where intended.

If the page still appears

Google lists what to try, in this order:

1. 1

   Confirm that Googlebot sees the directive

   The URL Inspection tool in Search Console shows the HTML Googlebot received. The tag or the header has to be in it.

2. 2

   Request a recrawl

   A change takes effect only after Google recrawls the page, and you can request a recrawl to start it sooner.

3. 3

   Allow time

   A recrawl can take from several days to several months.