noindex vs nosnippet for Google's AI answers

What does each rule do?

noindex tells Google not to show a page, media item or resource in search results. When Googlebot crawls a page and finds a noindex meta tag or header, Google drops the page from Search results entirely, whoever links to it. Because AI Overviews and AI Mode are Search features, and Google requires a page to be indexed to appear in them as a supporting link, noindex removes the page from those too.

nosnippet tells Google not to show a text snippet or video preview for a page in search results. The page stays indexed and can still rank; its listing just has no preview text. Google's documentation now adds the part that matters here: nosnippet will also prevent the content from being used as a direct input for AI Overviews and AI Mode.

What are the finer controls?

Two siblings of nosnippet let you be more precise. max-snippet sets a maximum number of characters for the text snippet, and Google says it will also limit how much of the content may be used as a direct input for AI Overviews and AI Mode. data-nosnippet is an HTML attribute you put on span, div and section elements to mark text that should not be used as a snippet, leaving the rest of the page usable.

• noindex

  Scope

  Whole page

  In Search

  Removed

  In AI Overviews and AI Mode

  Removed with Search

• nosnippet

  Scope

  Whole page

  In Search

  Listed, no preview text

  In AI Overviews and AI Mode

  Text not a direct input; no supporting link

• max-snippet

  Scope

  Whole page, by length

  In Search

  Preview capped

  In AI Overviews and AI Mode

  Limits how much text is a direct input

• data-nosnippet

  Scope

  Marked elements

  In Search

  Marked text left out of previews

  In AI Overviews and AI Mode

  Marked text left out

Google lists all four, nosnippet, data-nosnippet, max-snippet and noindex, as the ways to limit how a page appears in its AI features. It also says there are no additional requirements to appear in those features and no special files or schema needed, so these four are the whole control surface on Google's side.

When should you use noindex?

When a page should not be found in Google at all. Internal search results, thin tag and filter pages, staging copies, thank-you and confirmation pages, duplicate print versions. For these, removing them from Search is the point, and removing them from AI Overviews comes with it.

noindex is the wrong tool for a page you want found but not quoted. It removes the page from every Google surface, and there is no partial version.

When should you use nosnippet or its siblings?

When a page should stay findable but its text should not be reused. Paid content where the preview would give the answer away. Licensed text you are not allowed to have excerpted. Legal or regulatory wording that must not appear out of context. In each case nosnippet, or more often data-nosnippet on the sensitive sections, keeps the page in Search while keeping the protected text out of previews and, by Google's statement, out of direct use in AI Overviews and AI Mode.

The cost is real. A result without a snippet usually attracts fewer clicks, and a page whose text cannot be used is unlikely to be the supporting link in an AI Overview for that topic. Use the narrowest control that does the job: data-nosnippet on a paragraph before nosnippet on a page, max-snippet before either if the concern is only length.

Why are these rules often ignored?

Because robots.txt blocks the page. Google states that if a page is disallowed from crawling through robots.txt, any information about indexing or serving rules will not be found and will be ignored. The page can then stay indexed, from links, with no snippet control at all, because Google never read the rule. This is the most common mistake on this topic: a team adds noindex and a robots.txt disallow at the same time, and the second defeats the first.

# robots.txt: this group stops Googlebot reading the rule below
User-agent: *
Disallow: /search/

# in the head of /search/?q=pricing: never seen, so never applied
<meta name="robots" content="noindex">

The fix is ordering. To remove a page from Google, let Googlebot crawl it, serve the noindex rule, wait for the page to drop out, and only then consider whether a robots.txt rule is needed at all.

What did an answer engine tell people to use?

We track "how to block AI crawlers in robots.txt" every day. Perplexity's stored answer of 4 October 2026 closed with the right warning, that robots.txt is a polite instruction and not access control, and then advised "authentication and/or noindex" for private content. Half of that is sound. The other half needs the caveat above: if the same URL is also disallowed in robots.txt, Google never reads the noindex, and the page can stay in Search from links. The answer did not mention nosnippet at all, though it is the rule most sites asking that question actually want. That is the pattern across everything we track: in the 30 days to 4 October 2026, 30 of 4,740 ChatGPT and Perplexity answers told the reader about noindex, and none mentioned nosnippet, data-nosnippet or max-snippet. One ChatGPT answer of 30 September put it as a warning, "Don't accidentally use noindex"; no answer offered the finer rule.

Where the engine learned this shows in its sources. Across 30 Perplexity answers to the prompt in 30 days, the most-cited domains were searchenginejournal.com (20 of 290 citations, 6.9%), github.com and cyberciti.biz (19 each). Google's developer documentation, where both rules are defined, was not among Perplexity's six most-cited domains for the prompt. ChatGPT, asked the same question, gave developers.google.com 608 of 1,507 citations (40.4%). The same question, two engines, two very different reading lists.

When is the Search Console setting better than nosnippet?

When the decision is about the whole site, not a page. Google's Search Console now has a "Search generative AI" control, rolled out to all sites by 31 August 2026. Set to exclude, it keeps the site out of AI Overviews, AI Mode and AI Overviews in Discover, both as links and as grounding for the answer, while the site stays in ordinary results. Google says the change generally takes a few days to apply.

That makes it a third option beside the two rules. noindex leaves Search entirely, page by page. nosnippet and its siblings keep a page in Search and keep its text out of previews and AI answers, page by page or passage by passage. The setting is site-wide and all or nothing for AI features. A publisher that wants nothing quoted in AI answers uses the setting; a company that wants its pricing quoted but its licensed report protected uses data-nosnippet on the report and leaves the setting alone.

How do you set them on files that are not HTML?

With a header. Google states that any rule you can use in a robots meta tag can also be sent as an X-Robots-Tag HTTP response header. In the page and on the wire, the two look like this:

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

HTTP/1.1 200 OK
X-Robots-Tag: noindex

That is the only way to apply noindex or nosnippet to PDFs, images and other files that have no head element. data-nosnippet is the exception, because it marks parts of an HTML page and has no header form.

Do other answer engines follow these rules?

Their documentation does not say so. OpenAI's crawler page and Perplexity's describe robots.txt tokens, OAI-SearchBot and PerplexityBot, as the control for whether their search features can use a page. Neither mentions noindex or nosnippet. If you need a page kept out of ChatGPT search or Perplexity, disallow that path for their crawlers in robots.txt. If you need part of a page kept out of their answers, there is no documented equivalent of data-nosnippet; the honest options are to move the content behind authentication or accept that it can be quoted.

That asymmetry is worth writing down in your policy. A team that adds nosnippet to protect licensed text may believe it has protected it from all AI answers, when it has only addressed Google's.

How do you audit what your pages say?

Read the rules from the live pages, not from templates. Fetch each important URL, record the robots meta tag and any X-Robots-Tag header, and compare against robots.txt for the same path. Look for three patterns: a noindex on a page you want found, usually left over from a staging deploy; a nosnippet on a commercial page, usually added for one reason and forgotten; and a noindex on a page that robots.txt blocks, which does nothing.

Attensira's page inventory records each page's robots meta tag and canonical as part of reading the page, which is how this audit can run continuously instead of once. The next question is whether the pages you allow are the ones AI answers use, which only reading the answers can tell you.

What does a worked example look like?

Take a software company with a public pricing page, a gated report and a long legal page:

• Pricing

  Setting

  No noindex, no snippet limit

  Why

  It should be found and quoted; state the price in plain text near the top

• Gated report landing page

  Setting

  Indexed, data-nosnippet on the findings summary

  Why

  Findable, without giving the answer away in a preview

• Legal terms

  Setting

  Indexed, a low max-snippet

  Why

  Findable, not quoted at length

• Internal search results

  Setting

  noindex, and crawlable in robots.txt

  Why

  Out of Search, and Google can see the rule

Every one of those decisions is a page-level choice about Google. None of them changes what ChatGPT or Perplexity can read, which is decided in robots.txt.

How should you choose?

Start from the outcome. If the page should not be found in Google, use noindex and let it be crawled. If it should be found but not quoted, use the narrowest snippet control that protects what needs protecting. If it should be found and quoted, which is most commercial pages, use neither, and put the effort into making the quotable part worth quoting. Rules can only take a page out of an answer; nothing in a meta tag can put one in.