Nosnippet, Max-Snippet, Data-Nosnippet: The Only Granular AI Overview Controls

Operators asked Google for years for an AI Overviews opt-out switch, and Google pointed at the same four directives it documented for snippets — because in Google's model, AI features are snippet consumers. Since August 31, 2026 a site-level switch exists too, in Search Console, but these directives remain the only page- and passage-level controls. The robots meta tag reference says so explicitly, in the only granular AI-visibility language Google publishes. This page lays out each control verbatim, the AI-specific behavior, and the trade every one of them forces.

What are Google's four snippet controls?

Four documented, page-level directives govern how much of your content Google may display or reuse: nosnippet, max-snippet:[number], data-nosnippet, and noindex. The first two ship as robots meta tags or X-Robots-Tag headers, the third is an HTML attribute, and the fourth removes the page from results entirely. Per Google's robots meta tag reference, August 2026:

These four are the entire documented page-level control surface. Google's AI-features page names them as the way "to limit the information shown from your pages"; the one other AI control Google documents works per site, the Search generative AI setting in Search Console, rolled out to all sites on August 31, 2026 (Search Console Help).

The delivery method is worth separating from the directive. The same instruction can travel in the page's markup or in the HTTP response, and the header form — settable by a CDN, a proxy or a staging default that no page author ever reads — is defined at X-Robots-Tag.

How does nosnippet affect AI Overviews and AI Mode?

It is the complete shutter short of deindexing: Google states the content will not be used "as a direct input for AI Overviews and AI Mode," full stop. The page can stay indexed, can rank, and can still appear as a bare listing — Google notes a static image thumbnail "may still be visible" where it helps users — but your words stop feeding both the classic snippet and the generated answer.

That combination is the trade in its purest form. A nosnippet page competing on an informational query shows searchers a title with no supporting text, next to competitors offering full snippets and possible answer citations. For a site that earns from being read, that is close to unilateral disarmament, which is why the control exists for the opposite economics: content whose value is captured, not advertised, by extraction.

What do max-snippet's values actually mean?

max-snippet:[number] is a character budget for extraction, and its two special values mark the poles. 0 is documented as "equivalent to nosnippet" — no textual snippet, no direct AI input. -1 tells Google to "choose the snippet length that it believes is most effective to help users discover your content." Any positive integer caps the textual snippet at that many characters and, per the same reference, limits "how much of the content may be used as a direct input for AI Overviews and AI Mode."

Scope matters as much as the cap: Google documents the rule as applying to "all forms of search results," enumerating web search, Google Images, Discover, Assistant, AI Overviews, and AI Mode — 6 surfaces obeying 1 directive. A middle value like max-snippet:160 is sometimes proposed as a compromise, preserving classic-length snippets while limiting how much text AI features may lift. That reading is consistent with the documentation, but Google publishes no guidance on choosing a number — treat any specific threshold as untested, because it is.

Can data-nosnippet hide part of a page from AI Overviews?

Yes — it is the one control that works below page level. The attribute is boolean ("any value specified is ignored") and valid on span, div, and section elements; text inside is excluded from snippeting while everything outside remains eligible:

<p>This paragraph can be quoted in snippets and AI features.</p>
<div data-nosnippet>
  This block is excluded from snippet use.
</div>

The practical pattern for a site that wants visibility and containment at once: leave your definitions and answers extractable, and wrap the parts extraction would cannibalize — proprietary tables, member content teasers, the numbers you sell. The honest cost repeats at this scale too: shielded passages also vanish from classic snippets, so shield what converts on the page, never the text that wins the click.

Is there an AI-only opt-out that keeps normal snippets?

For a whole site, yes; for a single page or passage, no. Since August 31, 2026, Search Console's Search generative AI setting excludes a site's links and content from AI Overviews, AI Mode and Discover's generative AI features, and Google says the setting "isn't used as a ranking or inclusion signal affecting other parts of Search" (Search Console Help). Below site level, every documented lever restricts AI features and classic snippets together, and anyone claiming a markup trick that separates them is selling something undocumented.

So the decision has two parts. Leaving AI answers entirely is now a setting; keeping some pages or passages out while staying in is still "how much extraction, everywhere?" — a business-model question. A publisher watching answers absorb its traffic may rationally accept thinner classic snippets as the price of containment; the fuller version of that calculus, including the block-the-crawlers path, is in should publishers block or allow AI. Note the boundary with robots.txt controls: crawler tokens govern fetching, snippet controls govern display and reuse — and Google's AI training opt-out is a different token entirely.

Should you use snippet controls at all?

Only if extraction demonstrably costs you more than visibility earns — and for our own fleet the answer is no. Across our 3 production builds we run 0 snippet controls [our data]: our pages exist to be quoted, so every directive on this page points the wrong way for us, and the structural work goes into being worth extracting instead — the checklist in how to show up in AI Overviews.

Against our own interest, the honest scenario where we would reverse: if a site's revenue lives in the exact passages engines lift — paywalled analysis, priced data, comparison tables that are the product — data-nosnippet on those passages is the most precise instrument Google documents, and we would deploy it despite everything above. Just deploy it knowingly: fewer words offered to every surface, classic snippets included, verified by checking the live SERP afterward rather than assuming. How these controls fit the larger visibility system — crawlers, structure, measurement — is mapped in our generative engine optimization guide.